LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yoon, Deun Sol
Song, Hyung Seok
Lee, Kang Hoon
Lim, Woo Hyung
Abrégé
The present invention relates to a method and system for training an artificial intelligence model and provides a method and system for training an artificial intelligence model, which improve the prediction performance of an artificial intelligence model in various target tasks, the method comprising the steps of: acquiring input data comprising at least one constraint condition related to a target task; generating at least one candidate prediction value satisfying the at least one constraint condition, by processing the input data as an input into a pre-trained artificial intelligence model; calculating cost information corresponding to the at least one candidate prediction value; and training the artificial intelligence model by using the cost information as a reward, such that the artificial intelligence model outputs a prediction value that minimizes the cost information.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lim, Taeyoon
Ahn, Wonbin
Kim, Minjae
Lim, Woohyung
Hwang, Taewan
Lee, Gwangho
Kim, Euisoon
Cha, Jiwon
Abrégé
The present disclosure relates to a method and a system for automatically generating a portfolio analysis report. The method comprises: collecting structured data regarding configuration changes and operation performance of a target portfolio during a predetermined analysis target period; generating summary data by summarizing and classifying unstructured text data across a plurality of temporal intervals and subject areas by using a language model; generating a structured summary for each asset from public documents related to assets within the portfolio; constructing reference data from an official source; and then generating an integrated prompt in which a numerical signal section including the structured data and an event summary section including the summary data and the like are structurally divided, so that the language model generates a narrative that causally links a change in the portfolio and a cause thereof.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Taehyeon
Lee, Hyunsoo
Park, Sunghyun
Choi, Jihoon
Abrégé
The present disclosure relates to a patent review automation system using multi-agent discussion and dynamic search/re-ranking. The system breaks down a patent claim into atomic elements and ensures breakdown accuracy through original text substring verification. A plurality of search queries are generated for each broken down element to search for prior art, and a correspondence relationship is generated through multi-level re-ranking and an N×M coverage matrix. A first agent of a rejection perspective and a second agent of a permission perspective perform structured discussion, and a verification agent supervises the structured discussion and generates a component-specific decision. Dynamic evidence reinforcement and identification detection are performed during a discussion process. On the basis of a discussion result, a claim chart having a two-way analysis structure of novelty and non-obviousness is automatically generated, and a gap analysis-based improvement proposal is provided when rejected. The accuracy of analysis continuously improves through a human feedback loop.
ADAPTIVE ARTIFICIAL INTELLIGENCE SYSTEM AND METHOD THEREFOR FOR DYNAMIC CAUSAL STRUCTURE LEARNING OF TIME SERIES DATA AND COUNTERFACTUAL OUTCOME PREDICTION
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lyu, Juhyun
Kim, Jonghwan
Kim, Junghee
Lee, Sangmin
Yang, Jinseok
Jung, Hyemin
Lim, Woohyung
Abrégé
The present disclosure relates to an adaptive artificial intelligence system and a method therefor for dynamic causal structure learning of time series data and counterfactual outcome prediction, wherein the system automatically detects structural changes in data from non-stationary time series data in which underlying causal relationships change over time and segments the data into a plurality of regimes, dynamically learns a causal structure specialized for each regime through a bi-level optimization technique, and precisely predicts a probability distribution of potential outcomes according to a virtual intervention (i.e., a counterfactual query) of a user on the basis of the learned causal structure, and provides results accordingly.
AI AGENT-BASED BUSINESS PROCESS ANALYSIS AND AUTOMATION OPPORTUNITY DERIVATION METHOD, AND SYSTEM FOR PERFORMING SAME AND COMPUTER-READABLE RECORDING MEDIUM
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Choi, Jihoon
Abrégé
The present disclosure relates to a method and a system for analyzing a business process and deriving an automation opportunities by using an AI agent. The method structuralizes business data collected from a plurality of stakeholders and synthesizes same into a multi-dimensional integrated map (including time axis alignment, conflict and gap detection). Thereafter, pain points on the map are clustered on the basis of composite similarity so as to automatically generate AI solution questions, and an input idea is converted into a visual sketch. Finally, a process is included in which multi-axis feasibility is evaluated on the basis of data availability, AI technology reliability and business influence, and priority categories of solutions are classified.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
H04L 51/02 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel en utilisant des réactions automatiques ou la délégation par l’utilisateur, p. ex. des réponses automatiques ou des messages générés par un agent conversationnel
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Gunhee
Bae, Gyeongho
Abrégé
One embodiment provides a method comprising the steps of: receiving an image for analysis and storing same in at least one memory; loading the image for analysis from the at least one memory; extracting, by at least one processor, a vision feature map from the image for analysis; generating, by the at least one processor, a global token, a spatial token, and a temporal token for an image for training by using the vision feature map; receiving a user query for the image for analysis; generating, by the at least one processor, a text token by encoding the user query; encoding location information by applying harmonic rotation location embedding to an input sequence comprising the spatial token, the temporal token, and a text token; performing, on the input sequence comprising the tokens with the location information encoded, a predetermined attention operation having applied thereto a separation mask controlling interaction between the spatial token and the temporal token; generating a result of analysis of the image for analysis on the basis of a result of the attention operation; and inputting the result of the analysis into at least one subsequent processing component.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jeong, Han Seul
Park, Young Joon
Song, Hyung Seok
Lim, Woo Hyung
Abrégé
The present invention relates to a simulation method and system, and relates to a technology for providing a simulation method and system for implementing physical AI. The present invention provides a user interface for generating an environmental layout related to a robot operation, and causes the environmental layout generated by the user interface and a control strategy model to interlock with each other. The present invention is intended to execute robot simulation on the basis of the environmental layout and the control strategy model interlocked with each other. The executed result is output to the user interface.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Seo, Sanghyun
Kim, Jaeheon
Kang, Bumsoo
Jeon, Kijeong
Abrégé
The present disclosure relates to a method and a system for providing a follow-up question in real time through distributed computation between a server and a client in a conversational artificial intelligence service. The server derives a conversation purpose and related knowledge through chain-of-thought (CoT) inference based on a conversation history, so as to generate and rank follow-up question candidates, stores metadata including a KV cache pointer, and then transmits a candidate pack to the client. The client constructs a prefix trie index by using the received data, searches the index locally without server communication when a user inputs a keystroke, and provides a matching candidate in real time. Through this, computational load on the server is reduced and response latency is minimized.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Ki Young
Hwang, Do Yeong
Yim, Soo Rin
Park, Sung Joon
Lee, Kyung Wook
Abrégé
The present disclosure relates to a phenotype prediction system comprising: a memory storing one or more instructions; and at least one processor that executes the one or more instructions stored in the memory, wherein an operation performed by the one or more instructions comprises a step of learning phenotype prediction from omics data, and the step of learning the phenotype prediction comprises the steps of: acquiring at least one data including the omics data; generating an omics feature on the basis of the omics data; generating, by a reconstruction model, a latent vector on the basis of the omics feature; and predicting, by a phenotype prediction model, a phenotype on the basis of the latent vector.
G06N 3/0442 - Réseaux récurrents, p. ex. réseaux de Hopfield caractérisés par la présence de mémoire ou de portes, p. ex. mémoire longue à court terme [LSTM] ou unités récurrentes à porte [GRU]
G06N 3/044 - Réseaux récurrents, p. ex. réseaux de Hopfield
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Hwang, Do Yeong
Kim, Ki Young
Yim, Soo Rin
Park, Sung Joon
Lee, Kyung Wook
Abrégé
The present invention relates to a system and method for predicting a protein structure, wherein protein data including amino acid sequence information of one or more proteins is converted into embeddings, a first model and a second model each generate logits from the embeddings, a softmax function using a temperature parameter is applied to the logits respectively generated by the first model and the second model, to generate prediction distributions, a loss function is calculated by using the prediction distributions, and learning is carried out such that the loss function decreases.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Ki Young
Park, Sung Joon
Lee, Kyung Wook
Yim, Soo Rin
Hwang, Do Yeong
Abrégé
The present disclosure relates to a system and method for phenotype prediction, and provides a system and method for phenotype prediction, the method comprising a step of training phenotype prediction from omics data, wherein the step of training phenotype prediction comprises the steps of: acquiring omics data about a plurality of modalities from one or more omics datasets; converting the omics data about the plurality of modalities into modality-specific embeddings via an encoder; generating, by an integration model, an integrated embedding from the modality-specific embeddings; generating, by a prediction model, a phenotype prediction value from the integrated embedding; and improving phenotype prediction performance by calculating a loss function by using the phenotype prediction value and training to reduce the loss function.
G06N 3/0985 - Optimisation d’hyperparamètresMeta-apprentissageApprendre à apprendre
G16H 50/20 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour le diagnostic assisté par ordinateur, p. ex. basé sur des systèmes experts médicaux
G16H 50/30 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour le calcul des indices de santéTIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour l’évaluation des risques pour la santé d’une personne
G16B 15/30 - Ciblage de médicament à l’aide de données structurellesPrévision d’amarrage ou de liaison moléculaire
G06N 3/042 - Réseaux neuronaux fondés sur la connaissanceReprésentations logiques de réseaux neuronaux
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kyungeun
Cho, Hyeseung
Yoon, Suhee
Eo, Moonjung
Yoon, Sanghyu
Sim, Yeseul
Lim, Woohyung
Abrégé
An embodiment provides a method performed by a computer, the method comprising the steps of: receiving tabular data for analysis; generating, by at least one processor, analysis result information about the tabular data for analysis by using at least one artificial intelligence model, wherein the at least one artificial intelligence model is pre-trained on the basis of a training dataset comprising tabular augmented data generated on the basis of recognition of a user labeling pattern; and inputting the analysis result information into at least one subsequent processing component.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Ki Young
Yim, Soo Rin
Hwang, Do Yeong
Park, Sung Joon
Lee, Kyung Wook
Abrégé
The present disclosure relates to artificial intelligence and bioinformatics technology, and more specifically, to an artificial intelligence system and method for designing a protein binder capable of binding to a specific protein with high affinity and specificity by dynamically considering changes in protein structure.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jeong, Han Seul
Park, Young Joon
Song, Hyung Seok
Lim, Woo Hyung
Abrégé
The present invention relates to a method and a system for artificial intelligence inference, and provides a method and a system for artificial intelligence inference, which are for solving various problems through inference by using artificial intelligence.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kanghoon
Yoon, Deunsol
Hong, Sunghoon
Jung, Whiyoung
Park, Junseok
Kim, Geonhyeong
Lim, Woohyung
Lee, Soonyoung
Abrégé
The present disclosure relates to a method for automatically generating a robot control interface by means of a computer processor. Specifically, the present invention receives a request including a task description from a user, and generates structured metadata by loading hardware specifications and available variable information about a robot and a simulation environment. On the basis of this, a system prompt including an observation and behavior design area is configured, and then the system prompt is input to a pre-trained artificial intelligence model to automatically generate a candidate code set including observation and behavior model codes. A final execution environment wrapper for robot control learning is constructed by verifying the finally generated candidate code, and the corresponding state and result are implemented on a user interface.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kanghoon
Hong, Sunghoon
Park, Junseok
Yoon, Deunsol
Lim, Woohyung
Lee, Soonyoung
Jung, Whiyoung
Abrégé
The present disclosure relates to a method for high-speed prototyping and control of an industrial automation environment. When a user places infrastructure components through a visual editor and configures agent tasks, a system serializes corresponding placement information and operating parameters into a structured text format. Subsequently, a hardware specification file is parsed to normalize state variables and actuator spaces into metadata, the normalized metadata and task inputs are fed into a pre-trained artificial intelligence model to automatically generate interface code that mediates data exchange between agents and a simulation environment, and thus a behavior model is linked to the simulation environment on the basis of the generated code, and an interaction result of an agent cluster is visually displayed on a user interface. Through this, it is possible to rapidly establish an environment and intuitively pre-verify a control algorithm without complex hard-coding.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kanghoon
Hong, Sunghoon
Jung, Whiyoung
Yoon, Deunsol
Kim, Jeonghye
Shin, Yongjae
Jung, Suhyun
Yoo, Hyundam
Kim, Youngjin
Moon, Chanwoo
Lim, Woohyung
Lee, Soonyoung
Abrégé
The present invention relates to a method for controlling an artificial intelligence model based on multi-phase learning. An experiment setup is received from a user to load a benchmark dataset, and raw data is preprocessed according to a second-phase standard through an observation model function. First-phase learning is performed with the preprocessed data so as to generate a checkpoint, a specific parameter is transferred to a second-phase model and initialized, and then second-phase learning is performed through interacting with the environment. Finally, the learning lineage and connection relationship between the two phases are visualized and provided to a user interface. The present invention fundamentally prevents data standard mismatch between online and offline learning, and supports parameter transfer between heterogeneous algorithms, thereby enabling intuitive multi-phase learning monitoring and optimal phase switching.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jo, Ah Ra
Kim, Ji Ye
Chun, Se Hyun
Oh, Seung Yul
Jo, Yeon Sik
Lee, Seung Jun
Ryoo, Kwang Rok
Lee, Jong Min
Abrégé
The present invention relates to a method and system for document understanding, and the method for document understanding according to the present invention may comprise the steps of: receiving a user request including at least one document; identifying, through analysis of the document, at least one layout region each including at least one content having different data characteristics; specifying one or more models configured to perform processing on each of the at least one content, on the basis of the data characteristics of the content included in each of the at least one layout region; extracting the at least one content included in each of the at least one layout region by using each of the one or more models; converting the document into a form corresponding to the user request by using the at least one content extracted using the one or more models; and as a response to the user request, providing a user terminal with the document converted into the form corresponding to the user request.
G06V 30/412 - Analyse de mise en page de documents structurés avec des lignes imprimées ou des zones de saisie, p. ex. de formulaires ou de tableaux d’entreprise
G06F 40/166 - Édition, p. ex. insertion ou suppression
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jo, Ah Ra
Lee, Seung Jun
Ryoo, Kwang Rok
Abrégé
The present invention relates to a method and a system for document understanding. The document understanding method according to the present invention may comprise the steps of: specifying at least one document to be analyzed; inputting the at least one document into a document analysis model trained on the basis of a training document data set including at least one augmented document data; by analyzing the at least one document using the document analysis model, identifying at least one layout area having an attribute different from that of another layout area in the at least one document; extracting at least one content included in each of the identified at least one layout area; generating output data for the at least one document by using the extracted at least one content; and providing the output data to a user terminal.
G06V 30/412 - Analyse de mise en page de documents structurés avec des lignes imprimées ou des zones de saisie, p. ex. de formulaires ou de tableaux d’entreprise
G06V 30/413 - Classification de contenu, p. ex. de textes, de photographies ou de tableaux
G06V 30/19 - Reconnaissance utilisant des moyens électroniques
G16C 20/70 - Apprentissage automatique, exploration de données ou chimiométrie
G16C 20/40 - Recherche de structures chimiques ou de données physicochimiques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jo, Ah Ra
Kim, Ji Ye
Chun, Se Hyun
Jo, Yeon Sik
Lee, Seung Jun
Ryoo, Kwang Rok
Han, Hee Jae
Byun, Jae Seung
Do, Soo Jong
Moon, Chan Woo
Ham, Ji Won
Lee, Hae In
Oh, Seung Yul
Abrégé
The present invention relates to a method and system for document understanding, and provides a method and system for document understanding using a deep document understanding (DDU) technology.
G06V 30/19 - Reconnaissance utilisant des moyens électroniques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Chun, Se Hyun
Jo, Ah Ra
Jo, Yeon Sik
Oh, Seung Yul
Kim, Ji Ye
Abrégé
The present invention relates to a method and a system for document understanding. The document understanding method according to the present invention may comprise the steps of: specifying at least one document to be analyzed; processing the document to be analyzed, as an input of at least one information processing model configured to process information related to a chemical domain; generating, by using the at least one information processing model, a first processing result for molecular structure information included in the document to be analyzed and a second processing result for chemical reaction information included in the document to be analyzed; and generating output data for chemical information reflecting a correlation between the molecular structure information and the chemical reaction information by using the first processing result and the second processing result.
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
22.
SYSTEM AND METHOD FOR DETERMINING PROCESS SCHEDULING INFORMATION BY USING MULTI-AGENT
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yoo, Kyung Jae
Abrégé
One embodiment of the present disclosure relates to a method performed by a computer, the method comprising the operations of: generating, on the basis of an initial state of a process, a pivot schedule including an initial macro operation sequence; duplicating the pivot schedule so as to generate a plurality of branch schedules; using an artificial intelligence agent so as to reflect different operation scenarios or probabilistic behavior policies for the respective branch schedules and sequentially add macro operations, thereby expanding the schedule; evaluating fitness for the respective expanded branch schedules when a predefined synchronization point is reached; and, on the basis of the evaluation result, selecting at least one from among the plurality of branch schedules so as to repeatedly perform a process of updating the pivot schedule, thereby determining final scheduling information.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
G05B 19/418 - Commande totale d'usine, c.-à-d. commande centralisée de plusieurs machines, p. ex. commande numérique directe ou distribuée [DNC], systèmes d'ateliers flexibles [FMS], systèmes de fabrication intégrés [IMS], productique [CIM]
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jo, Ah Ra
Kim, Ji Ye
Chun, Se Hyun
Oh, Seung Yul
Jo, Yeon Sik
Lee, Seung Jun
Ryoo, Kwang Rok
Abrégé
The present invention relates to a method and system for document understanding, and the method for document understanding according to the present invention may comprise the steps of: receiving at least one document from a user terminal; identifying, through analysis of the document, at least one layout region each including at least one content having different data characteristics; specifying one or more models configured to perform processing on each of the at least one content, on the basis of the data characteristics of the content included in each of the at least one layout region; extracting the at least one content included in each of the at least one layout region by using each of the one or more models; generating a structured document understanding result related to the at least one content by performing multimodal data processing on the at least one content extracted using the one or more models and having the different data characteristics; and on the basis of information included in the structured document understanding result, providing an editing function for the at least one document to the user terminal.
G06V 30/412 - Analyse de mise en page de documents structurés avec des lignes imprimées ou des zones de saisie, p. ex. de formulaires ou de tableaux d’entreprise
G06F 40/166 - Édition, p. ex. insertion ou suppression
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kang Hoon
Hong, Sung Hoon
Yoon, Deun Sol
Jung, Whi Young
Park, Jun Seok
Lee, Jin Sang
Jung, Su Hyun
Yoo, Hyun Dam
Moon, Chan Woo
Lim, Woo Hyung
Lee, Soonyoung
Abrégé
A schedule providing method according to an embodiment of the present disclosure comprises the steps of: generating a first GUI for acquiring user input information; acquiring the user input information on the basis of the first GUI; specifying a target period; on the basis of the user input information and the target period, determining whether a schedule can be generated; generating the schedule during the target period when it is identified that the schedule can be generated; and controlling a display to output a second GUI for providing the schedule during the target period, wherein the schedule during the target period includes a first schedule pertaining to the operation of a warehousing tank during the target period, a second schedule pertaining to the operation of a blending tank during the target period, and a third schedule pertaining to the operation of a cracking furnace during the target period, and at least one among the first schedule, the second schedule, and the third schedule may be determined by one or more AI agents for performing reinforcement learning.
G05B 19/418 - Commande totale d'usine, c.-à-d. commande centralisée de plusieurs machines, p. ex. commande numérique directe ou distribuée [DNC], systèmes d'ateliers flexibles [FMS], systèmes de fabrication intégrés [IMS], productique [CIM]
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
G16C 20/10 - Analyse ou conception des réactions, des synthèses ou des procédés chimiques
G06N 3/10 - Interfaces, langages de programmation ou boîtes à outils de développement logiciel, p. ex. pour la simulation de réseaux neuronaux
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jeong, Dae Woong
Yang, Hong Jun
Ko, Seung Woo
Abrégé
A system for compressing molecular data according to the present invention comprises: a preprocessing unit for converting a plurality of pieces of original molecular data into molecular graphs, respectively; and a molecular compression unit which searches for a maximum common substructure between a first molecular graph and a second molecular graph among the molecular graphs, merges the first and second molecular graphs on the basis of the searched maximum common substructure to generate an integrated molecular graph, verifies whether each of atoms constituting the integrated molecular graph satisfies the valence rule, and tags each node of the integrated molecular graph for which verification has been completed with identifier information for identifying an original molecule from which the node is derived.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Rodrigo, Hormazabal
Jeong, Dae Woong
Han, Se Hui
Abrégé
A foundation model for performing molecular-level tasks by learning multimodal data in the form of one-dimensional text and two-dimensional graphs, and a system therefor, according to an embodiment of the present invention, enable various molecular-unit tasks such as chemical reaction prediction, molecular attribute prediction, and natural language description generation to be effectively processed through a single foundation model. In addition, it is possible to increase prediction accuracy of the model by maximizing utilization of two-dimensional molecular graph information, and automatically generate, on the basis of statistical sparsity, high-quality descriptive text that emphasizes core and distinctive features of each molecule.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Jae Hoon
Park, Sung Woo
Abrégé
The objective of one embodiment of the present disclosure is to provide a system comprising: one or more processors; an artificial intelligence prediction model; and one or more memories that collectively store instructions that, when executed by the one or more processors, instruct a computing system to perform operations. The operations may comprise: generating a plurality of observation data by sampling past data with an arbitrary time distribution; generating a propagation signal of each of the plurality of observation data based on mean field theory; determining a predicted value by aggregating propagation signal calculation results for the plurality of observation data; determining a loss value on the basis of the difference between the predicted value and an actual value; and training the artificial intelligence prediction model until the loss value is equal to or less than a preset value.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Jae Hoon
Park, Sung Woo
Abrégé
An embodiment of the present disclosure may provide a system comprising: at least one processor; and at least one memory that stores instructions, when executed by the at least one processor, causing the system to perform operations, wherein the operations comprise: an operation of acquiring an observation sequence; an operation of generating, with the observation sequence and time information as inputs, a latent sequence at a first time point by using an encoder module; an operation of generating a latent sequence at a future time point by inputting, into a first linear operation module, a first parameter regarding coefficients of a linear probability differential equation and the latent sequence at the first time point; and an operation of converting the latent sequence at the future time point into a prediction sequence and outputting same, by using a decoder module.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Jin Sik
Choi, Jung Kyu
Abrégé
The present disclosure relates to a method and a system for providing a multi-step inference artificial intelligence agent service. According to an embodiment, the system may dynamically reconfigure an agent persona on the basis of user meta information and decompose a query into a plurality of subtasks to establish an execution plan. In addition, a data analysis task is executed in a sandbox environment, and execution integrity can be ensured through a self-correction loop that corrects a code by itself when an error is detected. Furthermore, it is possible to maximize work automation efficiency by providing reliability-based search results and analysis content as artifacts in the form of editable native office objects.
G06N 3/063 - Réalisation physique, c.-à-d. mise en œuvre matérielle de réseaux neuronaux, de neurones ou de parties de neurone utilisant des moyens électroniques
G06N 3/084 - Rétropropagation, p. ex. suivant l’algorithme du gradient
METHOD FOR TRAINING GENERATIVE ARTIFICIAL INTELLIGENCE MODEL HAVING HYBRID ATTENTION STRUCTURE, AND METHOD AND SYSTEM FOR PROVIDING CONTENT ON BASIS OF INFERENCE CONTROL BY USING SAME
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Jin Sik
Kim, Yi Reun
Abrégé
The present disclosure relates to a method for training a generative artificial intelligence model having a hybrid attention architecture, and a system for providing content on the basis of inference control by using same. A method for providing content, according to one embodiment of the present disclosure, comprises the steps of: analyzing a query received from a user terminal, so as to determine an execution mode to be a non-inference mode or an inference mode; and, according to the determined mode, generating a response by using an artificial intelligence model having a hybrid attention mechanism. The artificial intelligence model has a hybrid structure in which a global attention layer and local attention layers are mixed and arranged in a preset ratio, and includes a rearrangement layer normalization structure, thereby ensuring both long context processing efficiency and learning stability. In addition, the model can be in a state of having been trained through integrated mode fine-tuning in which inference data and non-inference data are mixed and reinforcement learning based on global gain policy optimization.
G06N 3/063 - Réalisation physique, c.-à-d. mise en œuvre matérielle de réseaux neuronaux, de neurones ou de parties de neurone utilisant des moyens électroniques
G06N 3/084 - Rétropropagation, p. ex. suivant l’algorithme du gradient
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Jin Sik
Kim, Yi Reun
Abrégé
The present disclosure relates to an artificial intelligence model having a hybrid attention-based mixture-of-experts architecture and an operating method thereof. The operating method of an artificial intelligence model according to an embodiment of the present disclosure includes a step of converting input data into an embedding vector, and calculating the embedding vector through the artificial intelligence model to generate a final output. The artificial intelligence model includes a mixture-of-experts block group that selectively utilizes some of a plurality of experts, and a hybrid attention layer that combines and performs local attention and global attention. According to the present disclosure, computational efficiency and learning stability of a large-scale parameter model can be simultaneously secured through a mixture-of-experts network architecture of a sandwich structure, and computational complexity during long context processing can be significantly reduced through sliding window-based hybrid attention.
G06N 3/063 - Réalisation physique, c.-à-d. mise en œuvre matérielle de réseaux neuronaux, de neurones ou de parties de neurone utilisant des moyens électroniques
G06N 3/084 - Rétropropagation, p. ex. suivant l’algorithme du gradient
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jang, Jong Seong
Lee, Soon Young
Lee, Jun Hyun
Abrégé
An embodiment provides a method comprising the steps of: receiving an activation signal for accessing data in at least one memory; accessing a data structure in the at least one memory according to the reception of the activation signal, wherein the data structure includes a diagnostic image; loading the diagnostic image from the at least one memory; generating, by at least one processor, at least one response with respect to the diagnostic image by using at least one artificial intelligence model using the diagnostic image as an input, wherein the at least one artificial intelligence model is pre-trained to perform image analysis on the basis of a feature representation for a causal dependency relationship between a low-magnification feature and a high-magnification feature of the diagnostic image; and inputting (ingesting) the at least one response to at least one subsequent processing component.
G06T 3/4053 - Changement d'échelle d’images complètes ou de parties d’image, p. ex. agrandissement ou rétrécissement basé sur la super-résolution, c.-à-d. où la résolution de l’image obtenue est plus élevée que la résolution du capteur
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yun, Ju Seung
Jang, Jong Seong
Hu, Yi
Lee, Soon Young
Abrégé
An embodiment provides a method comprising the steps of: receiving an activation signal for accessing data in at least one memory; accessing a data structure in the at least one memory according to the reception of the activation signal; loading a diagnostic image from the at least one memory; generating, by at least one processor, at least one response with respect to the diagnostic image by using at least one artificial intelligence model using the diagnostic image as an input; and inputting the at least one response to at least one subsequent processing component.
G06T 3/4053 - Changement d'échelle d’images complètes ou de parties d’image, p. ex. agrandissement ou rétrécissement basé sur la super-résolution, c.-à-d. où la résolution de l’image obtenue est plus élevée que la résolution du capteur
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Byoung Jip
Jang, Young Soo
Logeswaran, Lajanugen
Kim, Geon Hyeong
Kim, Yu Jin
Lee, Hong Lak
Lee, Moon Tae
Abrégé
The present invention relates to a method and a system for providing an agent based on multi-language model inference. More specifically, the present invention relates to a method and a system for providing an agent based on multi-language model inference, and provides a method and a system for providing an agent based on multi-language model inference, in which a language model generates a plurality of inference paths and can select, from among same, a path having the highest probability of achieving a goal.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Song, Ho Sung
Lee, Kyung Jae
Shim, Dong Sub
Min, Kyung Koo
Park, Sung Hyun
Hwang, Da Sol
Abrégé
The present invention relates to a data generation system and a control method thereof, and provides: a data generation system for automatically generating data specialized for at least one domain using an artificial intelligence model; and a control method thereof.
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
G06F 18/214 - Génération de motifs d'entraînementProcédés de Bootstrapping, p. ex. ”bagging” ou ”boosting”
G06F 11/34 - Enregistrement ou évaluation statistique de l'activité du calculateur, p. ex. des interruptions ou des opérations d'entrée–sortie
G06F 16/3329 - Formulation de requêtes en langage naturel
G06F 16/338 - Présentation des résultats des requêtes
G06F 40/58 - Utilisation de traduction automatisée, p. ex. pour recherches multilingues, pour fournir aux dispositifs clients une traduction effectuée par le serveur ou pour la traduction en temps réel
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Song, Ho Sung
Lee, Kyung Jae
Shim, Dong Sub
Min, Kyung Koo
Abrégé
The present invention relates to a data generation system and a control method thereof, and provides a data generation system and a control method thereof, in which the data generation system automatically generates, using an artificial intelligence model, domain-specific data and data reflecting user preference.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Choi, Sung Ik
Lee, Moon Tae
Abrégé
The present invention relates to a method and system for image detection, and provides a method and system for image detection, which can distinguish between a real image and an artificially generated image by using an artificial intelligence model.
G06V 20/00 - ScènesÉléments spécifiques à la scène
G06V 40/40 - Détection d’usurpation, p. ex. détection d’activité
G06V 40/16 - Visages humains, p. ex. parties du visage, croquis ou expressions
G06V 10/774 - Génération d'ensembles de motifs de formationTraitement des caractéristiques d’images ou de vidéos dans les espaces de caractéristiquesDispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant l’intégration et la réduction de données, p. ex. analyse en composantes principales [PCA] ou analyse en composantes indépendantes [ ICA] ou cartes auto-organisatrices [SOM]Séparation aveugle de source méthodes de Bootstrap, p. ex. "bagging” ou “boosting”
G06V 10/44 - Extraction de caractéristiques locales par analyse des parties du motif, p. ex. par détection d’arêtes, de contours, de boucles, d’angles, de barres ou d’intersectionsAnalyse de connectivité, p. ex. de composantes connectées
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yoa, Seung Dong
Lee, Seung Jun
Cho, Hye Seung
Kim, Bum Soo
Lim, Woo Hyung
Abrégé
The present invention relates to an image recognition performance optimization method and system, and provides an image recognition performance optimization method and system, in which a visual token is re-tokenized in units of semantics in order to improve image recognition efficiency.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Cui, Run
Kim, Seung Hwan
Jeon, Gi Young
Kang, Byung Jun
Sim, Ye Seul
Yoa, Seung Dong
Cho, Hye Seung
Oh, Yeon Joo
Lee, Jin Sang
Jung, Su Hyun
Yoo, Hyun Dam
Abrégé
An embodiment of the present disclosure relates to a method and system for generating a training dataset through data consistency analysis and providing an artificial intelligence service by using same, wherein, in a training dataset environment that may include noise labels, feature consistency, structure consistency, and gradient consistency of data are analyzed in an integrated manner so as to detect and correct noise labels and suppress the influence thereof.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kang, Byung Jun
Lee, Soo Chan
Kim, Seung Hwan
Sim, Ye Seul
Yoa, Seung Dong
Cho, Hye Seung
Oh, Yeon Joo
Lee, Jin Sang
Jung, Su Hyun
Yoo, Hyun Dam
Cui, Run
Jeon, Gi Young
Abrégé
A Bayesian integrated vision inspection method and a system thereof according to one embodiment of the present disclosure implement a vision inspection framework which integrates and processes various types of data and labels by modeling class-specific distributions of patch embeddings extracted from vision data and classifying the patch embedding through inference using Bayes' theorem.
G06V 10/764 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant la classification, p. ex. des objets vidéo
G06F 18/2415 - Techniques de classification relatives au modèle de classification, p. ex. approches paramétriques ou non paramétriques basées sur des modèles paramétriques ou probabilistes, p. ex. basées sur un rapport de vraisemblance ou un taux de faux positifs par rapport à un taux de faux négatifs
G06N 3/06 - Réalisation physique, c.-à-d. mise en œuvre matérielle de réseaux neuronaux, de neurones ou de parties de neurone
G01N 21/88 - Recherche de la présence de criques, de défauts ou de souillures
G06T 7/80 - Analyse des images capturées pour déterminer les paramètres de caméra intrinsèques ou extrinsèques, c.-à-d. étalonnage de caméra
41.
VISION INSPECTION METHOD BASED ON COARSE-TO-FINE PATCH LEVEL CLASSIFICATION AND SYSTEM THEREOF
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kang, Byung Jun
Hyun, Jee Ho
Koh, Young San
Cui, Run
Jeon, Gi Young
Kim, Seung Hwan
Sim, Ye Seul
Yoa, Seung Dong
Cho, Hye Seung
Oh, Yeon Joo
Lee, Jin Sang
Jung, Su Hyun
Yoo, Hyun Dam
Abrégé
A vision inspection method based on coarse-to-fine patch level classification and a system thereof according to one embodiment of the present disclosure relate to a vision inspection method based on coarse-to-fine patch level classification and a system thereof which simultaneously improve the accuracy and processing speed of vision inspection by coarsely detecting a region suspected of being defective from given vision data and finely classifying the detected region on the basis of a pre-built data pool.
G06V 10/74 - Appariement de motifs d’image ou de vidéoMesures de proximité dans les espaces de caractéristiques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
G01N 21/88 - Recherche de la présence de criques, de défauts ou de souillures
G06V 10/40 - Extraction de caractéristiques d’images ou de vidéos
G06N 3/06 - Réalisation physique, c.-à-d. mise en œuvre matérielle de réseaux neuronaux, de neurones ou de parties de neurone
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kang, Byung Jun
Kim, Sang Yun
Kim, Seung Hwan
Sim, Ye Seul
Yoa, Seung Dong
Cho, Hye Seung
Oh, Yeon Joo
Lee, Jin Sang
Jung, Su Hyun
Yoo, Hyun Dam
Cui, Run
Jeon, Gi Young
Abrégé
A vision inspection camera parameter setting automation method and system according to an embodiment of the present disclosure relate to a vision inspection camera parameter setting automation method and system which quantitatively measure a distance between features extracted from vision data of good (OK) and defective (NG) samples, and automatically optimize parameters of a vision inspection camera by searching for a point that maximizes the measured distance.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Jae Kyeom
Sohn, Sung Ryull
Choi, Ji Hoon
Jo, Jeong Won
Lee, Hong Lak
Choi, Ye Muk
Abrégé
The present invention relates to a user-customized data compliance management method and system for: analyzing dataset license compliance by using an AI model; allowing a plurality of real or virtual agents to collaborate (discuss) and assess the analyzed result; and further refining data so as to satisfy a risk class not smaller than a preset threshold value.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Sim, Myo Seop
Min, Kyung Koo
Park, Min Jun
Jung, Hae Min
Abrégé
The present invention relates to a response generation method and system, and provides a response generation method and system using generative artificial intelligence (AI) or a large language model (LLM).
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Ji Won
Abrégé
A spatial information processing system according to an embodiment of the present invention may comprise the operations of: receiving an image and a query as inputs; extracting, from the image, a multi-dimensional visual feature representation including captions, object coordinates, and OCR text; converting the multi-dimensional visual feature representation into a linguistic representation understandable by a large-scale language model by using at least one lightweight artificial intelligence model; and receiving the converted linguistic representation and the query as inputs and generating a final response.
G06F 8/38 - Création ou génération de code source pour la mise en œuvre d'interfaces utilisateur
G06F 18/214 - Génération de motifs d'entraînementProcédés de Bootstrapping, p. ex. ”bagging” ou ”boosting”
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06F 16/532 - Formulation de requêtes, p. ex. de requêtes graphiques
G06F 16/538 - Présentation des résultats des requêtes
G06F 16/3329 - Formulation de requêtes en langage naturel
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Park, Young Yong
Kim, Ji Hoon
Paik, Young Min
Hwang, Tae Wan
Abrégé
An intelligent data analysis system according to an embodiment of the present invention may comprise the operations of: acquiring data to be analyzed and a user query in a natural language form; analyzing the user query so as to derive an analysis objective, and establishing an analysis plan defining a logical procedure to be performed in order to achieve the analysis objective; generating, on the basis of the established analysis plan and structural attribute information of the data to be analyzed, a program code mechanically executable in a computing environment; and driving the program code in an execution environment so as to calculate and output an analysis result corresponding to the user query.
G06F 8/38 - Création ou génération de code source pour la mise en œuvre d'interfaces utilisateur
G06F 18/214 - Génération de motifs d'entraînementProcédés de Bootstrapping, p. ex. ”bagging” ou ”boosting”
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06F 16/532 - Formulation de requêtes, p. ex. de requêtes graphiques
G06F 16/538 - Présentation des résultats des requêtes
G06F 16/3329 - Formulation de requêtes en langage naturel
47.
METHOD AND SYSTEM FOR ANALYZING IMAGE ON BASIS OF SEPARATE INTERPRETATION FOR SPATIAL FEATURE AND TEMPORAL FEATURE
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Gun Hee
Bae, Gyeong Ho
Abrégé
An embodiment provides a method comprising the steps of: receiving an image for analysis and storing the image in at least one memory; loading the image for analysis from the at least one memory; generating, by at least one processor, an analysis result for the image for analysis by using at least one artificial intelligence model using the image for analysis as an input thereto, wherein the at least one artificial intelligence model is pre-trained to output the analysis result for the image by performing a predetermined operation using a disentangling mask for controlling the interaction between a spatial token including spatial information on the image and a non-spatial token including temporal information or language information; and ingesting the analysis result into at least one subsequent processing component.
G06F 8/38 - Création ou génération de code source pour la mise en œuvre d'interfaces utilisateur
G06F 18/214 - Génération de motifs d'entraînementProcédés de Bootstrapping, p. ex. ”bagging” ou ”boosting”
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06F 16/532 - Formulation de requêtes, p. ex. de requêtes graphiques
G06F 16/538 - Présentation des résultats des requêtes
G06F 16/3329 - Formulation de requêtes en langage naturel
48.
METHOD AND SYSTEM FOR PREDICTING AND PROFILING IMPURITIES IN CHEMICAL REACTION
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yoo, In Wan
Abrégé
The method and system of the present invention, for predicting and profiling impurities in a chemical reaction, match reactants using reaction templates and thereby arrange impurity candidates in a priority queue. Provided is an efficient and accurate method for predicting impurities by combination of text conditions and reaction characteristics through text embedding and RXN embedding.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kang, Bum Soo
Seo, Sang Hyun
Kim, Jae Heon
Lee, Dam
Kim, Eui Soon
Jeon, Ki Jeong
Abrégé
According to an embodiment of the present invention, a prompt recommendation and judgment system may comprise: an operation of analyzing context of user input data and in which at least one artificial intelligence model generates a plurality of candidate prompts corresponding to the context; an operation in which the at least one artificial intelligence model selects a reference judgment index matching the context and creates a dynamic judgment rubric by extending the reference evaluation index so as to correspond to the characteristics of the context; and an operation of calculating suitability for the plurality of candidate prompts by applying the dynamic judgment rubric and determining a recommendation prompt on the basis of the calculated result.
G06F 8/38 - Création ou génération de code source pour la mise en œuvre d'interfaces utilisateur
G06F 18/214 - Génération de motifs d'entraînementProcédés de Bootstrapping, p. ex. ”bagging” ou ”boosting”
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06F 16/532 - Formulation de requêtes, p. ex. de requêtes graphiques
G06F 16/538 - Présentation des résultats des requêtes
G06F 16/3329 - Formulation de requêtes en langage naturel
50.
MULTI-AGENT REINFORCEMENT LEARNING SYSTEM, METHOD, AND APPARATUS
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jung, Jae Heon
Lee, Kang Hoon
Yoon, Deun Sol
Lee, Jin Sang
Moon, Chan Woo
Jang, Young Soo
Hong, Sung Hoon
Yoo, Hyun Dam
Ham, Ji Won
Kim, Jeong Hye
Shin, Yong Jae
Jung, Su Hyun
Kim, Geon Hyeong
Jung, Whi Young
Abrégé
Provided, in one embodiment of the present disclosure, is a method performed by a computer. The method may comprise the operations of: when first observation data of a first agent is received at a first time point, updating history information of the first agent by inputting the first observation data of the first agent and time information corresponding to the first observation data into a history encoder of the first agent; when observation data of a second agent is not received at the first time point, maintaining history information of the second agent; and calculating a value function by inputting the history information of the first agent and the history information of the second agent into an aggregation module.
G06N 3/0442 - Réseaux récurrents, p. ex. réseaux de Hopfield caractérisés par la présence de mémoire ou de portes, p. ex. mémoire longue à court terme [LSTM] ou unités récurrentes à porte [GRU]
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Sim, Ye Seul
Kim, Dong Min
Yoon, Su Hee
Yoon, Sang Hyu
Yoa, Seung Dong
Lee, Soon Young
Lim, Woo Hyung
Abrégé
An expert-agent interaction-based anomaly detection platform, and an operating method thereof, according to one embodiment of the present disclosure, which are capable of: continuously reflecting feedback of a domain expert in a system in a real-world industrial environment that dynamically changes, such as a smart factory; and interacting with a large-scale language model (LLM)-based agent so as to autonomously construct and update an anomaly detection model.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Sim, Ye Seul
Yoa, Seung Dong
Yoon, Su Hee
Yoon, Sang Hyu
Kim, Dong Min
Lim, Woo Hyung
Lee, Soon Young
Abrégé
A method for generating a dynamic anomaly detection workflow on the basis of agent orchestration and a modular protocol, and a system therefor according to an embodiment of the present disclosure relate to a method and a system therefor, the method autonomously generating and executing a customized anomaly detection workflow by allowing a large language model (LLM)-based agent to analyze characteristics of a user's natural language request or data in real time, and dynamically selecting and combining a plurality of functional tools defined by means of a model context protocol (MCP).
G06N 3/047 - Réseaux probabilistes ou stochastiques
53.
AGENT SYSTEM BASED ON MULTI-STAGE INFERENCE AND SELF-CORRECTION AND OPERATING METHOD THEREOF, AND SYSTEM FOR PROVIDING ARTIFICIAL INTELLIGENCE MODEL SERVICE PLATFORM INCLUDING AGENT SYSTEM AND OPERATING METHOD THEREOF
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Jin Sik
Kim, Yi Reun
Choi, Jung Kyu
Lee, Kyung Min
Jeon, Hyo Jin
Hwang, Jun Won
Yeen, Heui Yeen
Hong, Seok Hee
Kim, Sun Kyoung
Kim, So Yeon
Choi, Eun Bi
Abrégé
The present disclosure relates to an artificial intelligence agent system having multi-stage inference and self-correction functions and an integrated service platform hosting same. The agent system analyzes the complexity of a user query and deconstructs the user query into subtasks, establishes an execution plan based on data dependencies, and performs closed loop control for analyzing an error occurring during code execution in a sandbox environment and self-correcting the error. In addition, the platform system determines whether to approve distribution by preventing contamination of training data and performing automated red teaming using an adversarial prompt during registration of a tuned model. Furthermore, the platform performs precise rate limit control in units of tokens in response to an API request and generates differential charging data for assigning a weight to an output token, thereby ensuring efficient resource management and service stability.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
54.
ARTIFICIAL INTELLIGENCE-BASED GRAPH STRUCTURE LEARNING METHOD AND SYSTEM, AND GRAPH-BASED DATA PROCESSING METHOD AND SYSTEM USING SAME
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Seo Yoon
Jung, Hye Min
Lim, Woo Hyung
Abrégé
The present invention relates to an artificial intelligence-based graph structure learning method and system, and a graph-based data processing method and system using same, and provides an artificial intelligence-based graph structure learning method and system, and a graph-based data processing method and system using same, which are capable of more effectively analyzing and learning graph data.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Ahn, Won Bin
Kang, Dong Wan
Abrégé
The present invention relates to a method and a system for training a demand forecasting model. The method for training a demand forecasting model, according to the present invention, which is performed by a computer, comprises the steps of: specifying time series data of different attributes related to a future time point as training data for a demand forecasting target item; specifying static data related to a unique attribute of the demand forecasting target item as training data; processing the time series data to obtain at least one basis pattern data from the time series data; processing at least one of the static data or the time series data to generate at least one adjustment information to be applied to the basis pattern data; and training the demand forecasting model to calculate a demand forecasting value related to the demand forecasting target item by using a result of combining the basis pattern data and the adjustment information.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Oh, Seung Yul
Jo, Ah Ra
Lee, Jong Min
Kim, Kwang Min
Lee, Ye Jin
Jo, Yeon Sik
Abrégé
The present invention relates to a document understanding method and system, and provides a document understanding method and system using deep document understanding (DDU) technology.
G06V 30/19 - Reconnaissance utilisant des moyens électroniques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Hyun Min
Abrégé
One embodiment provides a method comprising the steps in which: an image to be analyzed is received and stored in at least one memory; the image to be analyzed is loaded from the at least one memory; at least one processor generates the result of the analysis of the image being analyzed using at least one artificial intelligence model having the image being analyzed as an input, the at least one artificial intelligence model being pre-trained to infer the geometric relationships between a plurality of objects in an image; and the result of the analysis is ingested into at least one post-processing component.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Sim, Myo Seop
Min, Kyung Koo
Park, Min Jun
Seong, Yong Ju
Abrégé
The present invention relates to a method and a system for generating training data, and provides a method and a system for generating training data, for training a model so as to be capable of self-debugging and correcting an execution error occurring in text-to-SQL generation.
G06F 11/3698 - Environnements pour l’analyse, le débogage ou le test de logiciel
G06F 11/14 - Détection ou correction d'erreur dans les données par redondance dans les opérations, p. ex. en utilisant différentes séquences d'opérations aboutissant au même résultat
G06F 16/3329 - Formulation de requêtes en langage naturel
G06F 16/338 - Présentation des résultats des requêtes
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yeen, Heui Yeen
Hong, Seok Hee
Yun, Hyeon Gu
Lee, Jin Sik
Abrégé
The present invention relates to a training data generation method and system, and provides a training data generation method and system in which a large-scale training data set for training an artificial intelligence model is generated.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Yoon, Hoon Sang
Lee, Won Kee
Cho, Il Min
Kim, Tak Young
Abrégé
The present invention relates to a method and a system utilizing a large language model for topic allocation based on quotations. According to the present invention, a computer-implemented method utilizing a large language model for topic allocation based on quotations may comprise the steps of: specifying content to be analyzed, which includes at least one sentence; specifying at least one topic related to the at least one sentence included in the content to be analyzed; extracting, by using a large language model, at least one quotation corresponding to each of the at least one topic from the content to be analyzed; and providing, by using the topic and the quotation, a topic analysis result for the content to be analyzed.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
An, So Young
Kim, Myoung Shin
Abrégé
The present invention relates to a red teaming assessment method and system for detecting vulnerabilities of an AI model on the basis of risk items and attack scenarios derived through ethics impact assessment, and provides the effect of enhancing the safety of artificial intelligence by retraining the model by utilizing, as feedback data, mitigation measures for the detected vulnerabilities.
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
62.
METHOD, DEVICE, AND SYSTEM FOR DETERMINING INFORMATION ON PROCESS COMPOSED OF MULTIPLE STEPS
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kang Hoon
Yoon, Deun Sol
Hong, Sung Hoon
Jung, Whi Young
Lim, Woo Hyung
Abrégé
One embodiment of the present disclosure may provide a system comprising one or more memories collectively storing instructions which, when executed by one or more processors, instruct the system to perform operations, the operations comprising: an operation of generating a plurality of groups including a first group and a second group; an operation of replicating a pivot schedule of each of the plurality of groups into a plurality of branch schedules for each of the plurality of groups; an operation of, for each of the plurality of branch schedules, expanding a schedule by performing a macro operation in which an agent trained on the basis of an artificial intelligence model reflects an operation scenario corresponding to each of the plurality of groups; an operation of evaluating a branch schedule of the first group and another group branch schedule at the same or higher level of constraint as the first group at a first synchronization time point with a suitability estimator network, and updating a pivot schedule of the first group according to the evaluation result; and an operation of determining a final schedule on the basis of the updated pivot schedule of the first group.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
63.
METHOD AND SYSTEM FOR GENERATING CODING GUIDE ON BASIS OF LARGE LANGUAGE MODEL
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Cho, In Hyuk
Oh, Whee Gun
Lee, Da Hyun
Lee, Hyun Soo
Abrégé
The present invention relates to a method and a system for generating a coding guide on the basis of a large language model. According to the present invention, the method for generating a coding guide on the basis of a large language model, which is performed by a computer, may comprise the steps of: receiving a user query related to at least one coding-related task among a plurality of different coding-related tasks from a user terminal; processing the user query as an input for a pre-trained large language model with respect to the plurality of coding-related tasks; obtaining, from the large language model, coding guide information which is related to at least one programming language related to the user query and is for the at least one coding task; generating a response to the user query for the at least one coding-related task by using the obtained coding guide information; and providing the response to the user query to the user terminal.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kang, Byung Jun
Lee, Soo Chan
Kim, Seung Hwan
Sim, Ye Seul
Yoa, Seung Dong
Cho, Hye Seung
Oh, Yeon Joo
Lee, Jin Sang
Jung, Su Hyun
Yoo, Hyun Dam
Cui, Run
Jeon, Gi Young
Abrégé
A vision inspection method using in-context learning, and a system therefor, according to one embodiment of the present disclosure, relate to a method and a system for performing, by using sequence model-based in-context learning, vision inspection on a query image by controlling an operation of a model through a prompt including a small number of images and label examples without updating model parameters.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Choi, Ji Hoon
Jang, Han Sol
Choi, Hye Won
Choi, Joo Young
Kim, Hyun
Jun, Chang Wook
Abrégé
The present invention relates to an artificial intelligence-based patent examination support method and system, which generate a search keyword by extracting components of a patent specification on the basis of various artificial intelligence models and algorithms, and automatically generate a report including the basis for determining patentability on the basis of the similarity with a prior art document that has been searched and filtered by using the search keyword.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Kang Hoon
Jung, Whi Young
Hong, Sung Hoon
Yoon, Deun Sol
Lim, Woo Hyung
Abrégé
One embodiment provides a system implemented by a computer, the system comprising at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the system to perform operations, wherein the operations include: collecting history data including information on past states and actions of each of a plurality of agents; encoding the history data with temporal information by using a history encoding module; integrating the encoded history data by applying a self-attention mechanism; estimating a joint value function on the basis of the integrated history data; and updating the policy of each of the plurality of agents on the basis of the joint value function.
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Hong Lak
Sohn, Sung Ryull
Choi, Ye Muk
Abrégé
The present disclosure provides an agentic AI driving method for dynamically determining an optimal toolchain on the basis of an execution mode according to a user input and an expected latency of a toolchain DB. Accordingly, an external tool and an internal module are controlled to execute a main task and output a final response including the execution result.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Chang Ho
Lee, Min Woo
Kang, Tae Gwan
Lee, Won Kee
Abrégé
A system, a method, and a program for acquiring data by using a language model are disclosed. The method comprises the steps of: receiving a query generated on the basis of a user query or a system event; analyzing the type of the query through a first language model; selecting, through the first language model, at least one API and a search range for acquiring real-time information according to the analyzed type of the query; requesting real-time data from a server by using the selected API and acquiring the real-time data as an API response; and generating and outputting a natural language response through a second language model by using the API response.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Ahn, You Bin
Won, Seung Pil
Lee, Do Haeng
Shin, Joong Bo
Abrégé
A table-type data analysis system, method, and program using a language model are disclosed. The method comprises the steps of: receiving a query and at least one table-type data, and converting the table-type data into at least one text-type data; inputting the query and the text-type data into a first language model to identify the text-type data corresponding to an analysis request included in the query and extracting the identified text-type data as analysis data; inputting the query and the analysis data into a second language model to generate an execution command for performing the analysis request; inputting the execution command into a third language model to convert the execution command into a code configured to perform the analysis request; and executing the code to output a data analysis result.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Ji Ye
Jo, Ah Ra
Chun, Se Hyun
Oh, Seung Yul
Jo, Yeon Sik
Abrégé
The present invention relates to a method and system for document understanding, and provides a method and system for document understanding using deep document understanding (DDU) technology.
G06V 30/19 - Reconnaissance utilisant des moyens électroniques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
G16C 20/30 - Prévision des propriétés des composés, des compositions ou des mélanges chimiques
G16C 20/10 - Analyse ou conception des réactions, des synthèses ou des procédés chimiques
G16C 20/90 - Langages de programmationArchitectures informatiquesSystèmes de bases de donnéesStockage de données
71.
METHOD AND SYSTEM FOR GENERATING RESPONSE BASED ON ARTIFICIAL INTELLIGENCE MODEL THROUGH EXTERNAL KNOWLEDGE RETRIEVAL TO WHICH USER FEEDBACK IS APPLIED
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Ham, Ji Won
Kim, Hee Jung
Jung, Su Hyun
Abrégé
One embodiment provides a method comprising the steps of: receiving a user query and storing same in at least one memory; loading the user query from the at least one memory; preprocessing the user query by at least one processor; generating, by the at least one processor, at least one response to the user query by using at least one artificial intelligence model taking, as an input, a prompt based on the preprocessed user query, wherein the prompt is generated on the basis of data obtained through external knowledge retrieval and is dynamically changed through user feedback; and ingesting the at least one response to at least one post-processing component.
G06F 16/3329 - Formulation de requêtes en langage naturel
G06F 16/338 - Présentation des résultats des requêtes
G06F 16/38 - Recherche caractérisée par l’utilisation de métadonnées, p. ex. de métadonnées ne provenant pas du contenu ou de métadonnées générées manuellement
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Suh, Min Kook
Eo, Moon Jung
Sim, Ye Seul
Lim, Woo Hyung
Abrégé
The present invention relates to a method and a system for data normalization for training using tabular data, and provides a B-spline-based method and system for data normalization for efficient deep neural network (DNN) training using tabular data.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Eo, Moon Jung
Lim, Tae Yoon
Oh, Yeon Ju
Suh, Min Kook
Lim, Woo Hyung
Abrégé
The present invention relates to a method and a system for combinatorial optimization. More particularly, the present disclosure relates to a method and a system for combinatorial optimization, in which parameter optimization is performed through a combination of a hybrid evolutionary algorithm (EA) and Bayesian optimization (BO).
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Eo, Moon Jung
Lee, Kyung Eun
Suh, Min Kook
Cho, Hye Seung
Sim, Ye Seul
Lim, Woo Hyung
Abrégé
The present invention relates to a method and system for data augmentation for learning tabular data. More specifically, the present invention relates to a method and system for data augmentation based on a self-attention mechanism for contrastive learning of tabular data.
G05B 13/02 - Systèmes de commande adaptatifs, c.-à-d. systèmes se réglant eux-mêmes automatiquement pour obtenir un rendement optimal suivant un critère prédéterminé électriques
75.
METHOD AND SYSTEM FOR PROVIDING DUAL LANGUAGE MODEL BASED ON LICENSE-FREE DATA
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Sohn, Sung Ryull
Kim, Jae Kyeom
Lee, Hong Lak
Jo, Jeong Won
Choi, Ji Hoon
Abrégé
The present invention relates to a method comprising: verifying a license of conversation data; using the conversation data as seed data to perform fine-tuning and additional training on a dual language model; constructing a license-free database based on the trained model; and providing the dual language model based on the constructed license-free database and services related thereto.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lim, Tae Yoon
Eo, Moon Jung
Oh, Yeon Ju
Suh, Min Kook
Lim, Woo Hyung
Abrégé
The present invention relates to a method and system for Bayesian optimization. More specifically, the present invention relates to a method and system for Bayesian optimization using individual parameter kernel density estimation (KDE) in a mixed variable environment.
G01N 21/88 - Recherche de la présence de criques, de défauts ou de souillures
G06F 30/27 - Optimisation, vérification ou simulation de l’objet conçu utilisant l’apprentissage automatique, p. ex. l’intelligence artificielle, les réseaux neuronaux, les machines à support de vecteur [MSV] ou l’apprentissage d’un modèle
G06F 18/2321 - Techniques non hiérarchiques en utilisant les statistiques ou l'optimisation des fonctions, p. ex. modélisation des fonctions de densité de probabilité
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
G06N 20/20 - Techniques d’ensemble en apprentissage automatique
77.
METHOD AND SYSTEM FOR PROVIDING WEB NAVIGATION SERVICE ON BASIS OF ACTION AGENT
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Hong Lak
Sohn, Sung Ryull
Abrégé
The present invention relates to a method and system for providing a web navigation service on the basis of an action agent , in which an AI model is trained with demonstration data in which actual web page manipulation is recorded, an action plan to be automatically performed by the trained AI is generated, and a web page is actually manipulated according to the generated plan to complete a requested task.
G06F 16/9038 - Présentation des résultats des requêtes
G06F 11/34 - Enregistrement ou évaluation statistique de l'activité du calculateur, p. ex. des interruptions ou des opérations d'entrée–sortie
G06V 30/414 - Extraction de la structure géométrique, p. ex. arborescenceDécoupage en blocs, p. ex. boîtes englobantes pour les éléments graphiques ou textuels
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Sohn, Sung Ryull
Kim, Jae Kyeom
Lee, Hong Lak
Jo, Jeong Won
Choi, Ji Hoon
Abrégé
The present invention relates to a technology for identifying a plurality of interactive elements on a web page when a user sets a goal, automatically generating an execution plan including at least one action to achieve the goal by identifying attributes of the identified interactive elements, and performing web navigation accordingly.
G06F 16/9038 - Présentation des résultats des requêtes
G06F 11/34 - Enregistrement ou évaluation statistique de l'activité du calculateur, p. ex. des interruptions ou des opérations d'entrée–sortie
G06V 30/414 - Extraction de la structure géométrique, p. ex. arborescenceDécoupage en blocs, p. ex. boîtes englobantes pour les éléments graphiques ou textuels
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Min Gyu
Kim, Min Seok
Byeon, Joon Hyeong
Lee, Jun Gi
Abrégé
The present invention relates to a method and a system for predicting process defects. More specifically, the present invention relates to a method and a system for predicting process defects capable of determining whether a product is defective by using time series data extraction.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Sang Woo
Yi, Sang Jun
Chang, Jun Bo
Jung, Hwan Seung
Choi, Jun Yong
Abrégé
The present invention relates to a method and system for predicting process defects. Provided are a method and system for predicting process defects capable of determining whether a product is abnormal by constructing an artificial intelligence model applicable to various process environments.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Eo, Moon Jung
Sim, Ye Seul
Lim, Woo Hyung
Cho, Hye Seung
Yoon, Su Hee
Yoon, Sang Hyu
Lee, Kyung Eun
Abrégé
One embodiment provides a method executed by a computer, the method comprising steps in which: at least one processor performs limited-range augmentation on numerical variable values included in original tabular data, so as to generate augmented tabular data; at least one artificial intelligence model is trained to recognize patterns of the original tabular data and the augmented tabular data; labeled tabular data stored in at least one memory is loaded; and the at least one artificial intelligence model is fine-tuned to output a result value for the tabular data by using the labeled tabular data.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Cui, Run
Kim, Seung Hwan
Hyun, Jee Ho
Jeon, Gi Young
Kang, Byung Jun
Kim, Sang Yun
Koh, Young San
Abrégé
A suspected defect region detection method according to one embodiment of the present disclosure is a method by which a computing system comprising a memory and a processor detects a suspected defect region, the method comprising the steps of: acquiring N (N>1) non-defective product images; acquiring an inspection target image, which is an image obtained by photographing an object for which the presence or absence of a defect is to be determined; aligning the inspection target image on the basis of the non-defective product images; dividing the aligned inspection target image into patch units; using a predetermined pre-training model so as to acquire a multi-scale feature based on the divided patches; estimating a non-defective product probability for each patch on the basis of the acquired multi-scale feature; determining, on the basis of the estimated non-defective product probability for each patch, whether a defect is suspected for the inspection target image; and providing a result of the suspected defect determination.
G06T 7/33 - Détermination des paramètres de transformation pour l'alignement des images, c.-à-d. recalage des images utilisant des procédés basés sur les caractéristiques
G06T 3/40 - Changement d'échelle d’images complètes ou de parties d’image, p. ex. agrandissement ou rétrécissement
G06V 10/74 - Appariement de motifs d’image ou de vidéoMesures de proximité dans les espaces de caractéristiques
G06T 3/18 - Déformation d’images, p. ex. réarrangement de pixels individuellement
G01N 21/88 - Recherche de la présence de criques, de défauts ou de souillures
83.
METHOD AND SYSTEM FOR DATA AUGMENTATION FOR LEARNING FROM TABULAR DATA
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Eo, Moon Jung
Lee, Kyung Eun
Cho, Hye Seung
Kim, Dong Min
Sim, Ye Seul
Lim, Woo Hyung
Abrégé
The present invention relates to a method and a system for data augmentation for learning from tabular data. More specifically, the present invention relates to a method and a system for expression level data augmentation for improving the performance of self-supervised learning (SSL) in tabular data.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kang, Byung Jun
Hyun, Jee Ho
Kim, Sang Yun
Koh, Young San
Cui, Run
Jeon, Gi Young
Kim, Seung Hwan
Abrégé
A knowledge distillation-based anomaly detection learning method according to an embodiment of the present invention comprises the steps of: acquiring a plurality of test sample images; calculating an anomaly score that is data obtained by digitizing a probability of a defect per patch feature for each of the plurality of test sample images; determining, as a defective candidate image, a test sample image in which the anomaly score satisfies a predetermined first condition; detecting, as a defective candidate patch feature, a patch feature in which the anomaly score satisfies a predetermined second condition, from among patch features of the defective candidate image; training a first adapter that is in charge of training in the current period, on the basis of the defective candidate patch feature; and updating a memory bank storing training information about a predetermined outlier detection model in a past period, on the basis of the training, wherein the first adapter is a network that performs the training in the current period, which reflects new information according to the plurality of test sample images while maintaining the training information in the past period at a certain level.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Cui, Run
Kim, Seung Hwan
Hyun, Jee Ho
Jeon, Gi Young
Kang, Byung Jun
Kim, Sang Yun
Koh, Young San
Abrégé
A deep learning-based environmentally adaptive outlier detection method according to an embodiment of the present invention is a method by which a computing system comprising a memory and a processor performs deep learning-based environmentally adaptive outlier detection and comprises the steps of: acquiring a corresponding fair quality image dataset for each predetermined material; acquiring corresponding Gaussian application information for each material, on the basis of Gaussian density estimation based on the fair quality image dataset; constructing a database comprising the Gaussian application information; determining whether a predetermined inspection target image is suspected of being defective, on the basis of the database; and providing a result of whether the predetermined inspection target image is suspected of being defective, wherein the Gaussian application information is information comprising mean (μ) and covariance (Σ) data of feature vectors extracted from N (N > 1) fair quality images included in the fair quality image dataset.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Park, Chang Young
Lee, Jae Wan
Yang, Hong Jun
Han, Se Hui
Abrégé
The artificial intelligence system of the present invention may comprise the steps of: calculating importance scores for data points of a dataset; calculating an oversampling rate for each of the data points; calculating sample weights on the basis of the oversampling rate for each of the data points; and oversampling the data points in response to the calculated sample weights.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Seo, Jun
Lim, Woo Hyung
Yang, Jin Seok
Choe, Hyeok Jun
Park, So Yeon
Kang, Dong Wan
Bae, Seo Hui
Abrégé
A time series prediction method using external information-based dynamic model control according to the present disclosure is a prediction method performed by a computer, and comprises the steps of: collecting time series data for at least one of a target to be predicted and a target influence variable associated with the target; obtaining unstructured data for at least one of the target and the target influence variable; extracting content information related to the target or the target influence variable from the obtained unstructured data; converting the extracted content information into an external control signal; and performing control to input the collected at least one time series data into a prediction model and to output a prediction result for the target on the basis of the converted external control signal.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Jeong Hye
Shin, Yong Jae
Lee, Kang Hoon
Jung, Whi Young
Hong, Sung Hoon
Yoon, Deun Sol
Lim, Woo Hyung
Abrégé
One embodiment of the present disclosure provides a reinforcement learning system comprising one or more processors, and one or more memories collectively storing instructions that, when executed by the one or more processors, cause the system to perform operations for performing offline reinforcement learning, and performing online reinforcement learning, wherein the operation for performing the offline reinforcement learning includes the operations for identifying a data retention area and a data non-retention area, and reducing a Q-value estimated in the data non-retention area.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lim, Woo Hyung
Lee, Kang Hoon
Yoon, Deun Sol
Hong, Sung Hoon
Kim, Jeong Hye
Shin, Yong Jae
Jung, Whi Young
Abrégé
An embodiment of the present disclosure may provide a system comprising: one or more processors; and one or more memories that collectively store instructions that, when executed by the one or more processors, cause the system to perform operations. In an embodiment, the operations may comprise the operations of: learning a first value function by performing offline reinforcement learning on the basis of offline data; learning a second value function by performing online pre-training; and performing online reinforcement learning by using a third value function generated on the basis of the first value function and the second value function.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Park, Chang Young
Yang, Hong Jun
Lee, Jae Wan
Han, Se Hui
Abrégé
A system for predicting physical properties of a material of the present invention may: extract a graph embedding by inputting material information to a first AI model; extract a text embedding by inputting a text description of a crystal structure of the material to a second AI model; divide the text embedding into a plurality of structural information embeddings; and combine the text embedding with at least one of the graph embedding or the plurality of structural information embeddings, wherein the structural information embeddings may be divided into global information, semi-global information, and local information about the crystal structure.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Park, Changyoung
Lee, Jaewan
Mrigi, Munjal
Yang, Hongjun
Han, Sehui
Abrégé
A system for predicting the physical properties of a material, of the present invention: extracts a graph embedding by inputting material information into a first AI model; extracts a text embedding by inputting a text description of a crystal structure of the material into a second AI model; partitions the text embedding into a plurality of structural information embeddings; and combines the graph embedding with at least one of the plurality of structural information embeddings, wherein the structural information embeddings may be partitioned by including global information, semi-global information, and local information of the crystal structure.
G06N 3/0895 - Apprentissage faiblement supervisé, p. ex. apprentissage semi-supervisé ou auto-supervisé
G16C 20/30 - Prévision des propriétés des composés, des compositions ou des mélanges chimiques
G16C 20/70 - Apprentissage automatique, exploration de données ou chimiométrie
G16C 60/00 - Science informatique des matériaux, c.-à-d. TIC spécialement adaptées à la recherche des propriétés physiques ou chimiques de matériaux ou de phénomènes associés à leur conception, synthèse, traitement, caractérisation ou utilisation
G16C 20/40 - Recherche de structures chimiques ou de données physicochimiques
92.
SYSTEM AND METHOD FOR GENERATING PROTEIN SEQUENCES FOR THERAPEUTIC AGENT IN CONSIDERATION OF HUMAN SAFETY
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Kiyoung
Yim, Soorin
Hwang, Doyeong
Abrégé
The present invention relates to a system and a method for generating protein sequence information. The system comprises: a memory for storing one or more instructions; and at least one processor for executing the one or more instructions stored in the memory, wherein an operation performed by the one or more instructions comprises the steps of: acquiring reference protein sequence information; repeatedly adding noise to the reference protein sequence information to generate protein sequence information to which noise has been added; and generating output protein sequence information by removing noise from the protein sequence information to which noise has been added, wherein the step of generating the output protein sequence information comprises repeatedly performing all or some of the steps of: removing noise from input protein sequence information according to at least one protein structure guidance specified by a user to generate protein sequence information from which noise has been removed; and adding noise to the protein sequence information from which noise has been removed according to at least one protein sequence guidance specified by the user, thereby generating protein sequence information to which noise has been added.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Geon Hyeong
Jang, Young Soo
Kim, Yu Jin
Kim, Byoung Jip
Lee, Hong Lak
Bae, Kyung Hoon
Lee, Moon Tae
Abrégé
The present invention relates to a large language model optimization method and system, and provides a large language model optimization method and system capable of improving the safety of a large language model. In particular, the present invention provides a large language model optimization method and system capable of generating a response that simultaneously satisfies helpfulness and safety.
METHOD FOR TRAINING HIGH-RESOLUTION IMAGE ANALYSIS MODEL BASED ON HIERARCHICAL PATCH SELECTION, AND METHOD AND SYSTEM FOR PROVIDING DIAGNOSTIC RESPONSE BASED ON HIGH-RESOLUTION IMAGE ANALYSIS
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jang, Jong Seong
Kim, Bum Soo
Lee, Soon Young
Lee, Jun Hyun
Abrégé
One embodiment provides a method for training a time series analysis model for a plurality of high-resolution images, performed by a computing device, the method for training a time series analysis model for a plurality of high-resolution images comprising the steps of: receiving a plurality of high-resolution images captured at a plurality of time points; outputting a plurality of analysis prediction values for the plurality of high-resolution images by performing an image analysis for each of the plurality of high-resolution images; configuring the plurality of analysis prediction values into time series data; and training a time series analysis model so as to output a predictive diagnostic response for the plurality of high-resolution images by having the time series data based on the plurality of analysis prediction values as input.
G16H 30/40 - TIC spécialement adaptées au maniement ou au traitement d’images médicales pour le traitement d’images médicales, p. ex. l’édition
G16H 30/20 - TIC spécialement adaptées au maniement ou au traitement d’images médicales pour le maniement d’images médicales, p. ex. DICOM, HL7 ou PACS
G06T 5/60 - Amélioration ou restauration d'image utilisant l’apprentissage automatique, p. ex. les réseaux neuronaux
95.
METHOD AND SYSTEM FOR LARGE LANGUAGE MODEL UNLEARNING, AND CONTROL METHOD FOR LARGE LANGUAGE MODEL UNLEARNING SYSTEM
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Cho, Sung Jun
Hwang, Da Sol
Lee, Moon Tae
Abrégé
The present invention relates to a method and a system for large language model unlearning, and a control method for the large language model unlearning system, which specify and compare forgetting data and retaining data for a pre-trained large language model and then measure the importance of parameters for each of the forgetting data and the retaining data by using a Fisher information matrix, so as to specify specific high-importance parameters, and initialize low-rank adaptation (LoRA) weights on the basis of the specific parameters such that the large language model to which the LoRA weights are applied can be optimized, thereby efficiently removing data to be eliminated, such as sensitive data, without affecting the inference and generation capability of the large language model.
METHOD FOR TRAINING HIGH-RESOLUTION IMAGE ANALYSIS MODEL ON BASIS OF PATCH SELECTION USING PROTOTYPE, AND METHOD AND SYSTEM FOR PROVIDING DIAGNOSTIC RESPONSE ON BASIS OF HIGH-RESOLUTION IMAGE ANALYSIS
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Choi, Hwan Il
Jang, Jong Seong
Kim, Bum Soo
Lee, Soon Young
Abrégé
One embodiment provides a method for training an image analysis model, comprising the steps of: generating a plurality of patch embeddings from an image; calculating similarity between a plurality of prototype embeddings and the plurality of patch embeddings; selecting a portion of the plurality of patch embeddings on the basis of the similarity for each prototype embedding; calculating a prediction value on the basis of the plurality of selected patch embeddings; and updating parameters of an image analysis model so that a loss based on the prediction value and correct answer data is minimized.
G06T 7/174 - DécoupageDétection de bords impliquant l'utilisation de plusieurs images
G16H 50/20 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour le diagnostic assisté par ordinateur, p. ex. basé sur des systèmes experts médicaux
G06V 20/69 - Objets microscopiques, p. ex. cellules biologiques ou pièces cellulaires
97.
GUIDE PROVIDING METHOD AND SYSTEM FOR SUPPORTING PERFORMANCE IMPROVEMENT OF MULTITASKING MODEL, DATA SAMPLING METHOD FOR GENERAL-PURPOSE MULTITASKING MODEL, AND METHOD AND SYSTEM FOR PROVIDING GENERAL-PURPOSE MULTITASKING MODEL INCLUDING SAME
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jeong, Dae Woong
Ko, Sung Moon
Lee, Su Min
Kim, Hyun Seung
Yim, Soo Rin
Abrégé
One embodiment of the present disclosure comprises the steps of: acquiring first input information specifying a predetermined domain characteristic; acquiring a validity index which is data quantitatively specifying a generation difficulty level for output data of a multitasking model according to the acquired first input information; generating first guide information which is data specifying the domain characteristic that lowers the generation difficulty level on the basis of the acquired validity index; providing the generated first guide information; acquiring input information specifying a predetermined domain characteristic; converting the acquired input information into a low-dimensional latent variable which is data expressed in a low-dimensional Gaussian space; acquiring a sampled latent variable which is data sampled from the converted low-dimensional latent variable; acquiring an optimized latent variable which is data optimized from the acquired sampled latent variable through a genetic algorithm; restoring the acquired optimized latent variable into a high-dimensional latent variable which is data expressed in a high-dimensional space; and providing the restored high-dimensional latent variable to the multitasking model.
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Kim, Min Jae
Lim, Tae Yoon
Kim, Jung Hee
Ahn, Won Bin
Abrégé
An asset portfolio prediction method according to one disclosure of the present invention is a method for predicting an asset portfolio performed by a computing system, the method comprising the steps of: determining a target prediction task which is data specifying a target task; collecting raw data for performing the determined target prediction task; arranging the collected raw data in time series according to a specific time axis; training a predetermined time series prediction model on the basis of the arranged raw data and the determined target prediction task; predicting portfolio asset weight data for each date for a predetermined prediction unit period on the basis of the trained time series prediction model; generating single portfolio asset weight data by processing the predicted portfolio asset weight data for each date at every predetermined rebalancing time point; and providing the generated single portfolio asset weight data.
G06N 3/049 - Réseaux neuronaux temporels, p. ex. éléments à retard, neurones oscillants ou entrées impulsionnelles
99.
METHOD FOR TRAINING HIGH-RESOLUTION IMAGE ANALYSIS MODEL, AND METHOD AND SYSTEM FOR PROVIDING DIAGNOSTIC ANSWERS BASED ON HIGH-RESOLUTION IMAGE ANALYSIS
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Lee, Min Soo
Yun, Ju Seung
Jang, Jong Seong
Kim, Jin Hyung
Kim, Sang Wook
Kim, Jong Hyun
Abrégé
Provided according to an embodiment is a method for training a high-resolution image analysis model, the method comprising the steps of: generating an integrated vision embedding from a high-resolution image by using a vision encoding module; concatenating the integrated vision embedding with a captioning text embedding corresponding to captioning information for the high-resolution image, so as to generate a first vision-language embedding; carrying out first training involving training a large language model (LLM) to output the captioning information by using the first vision-language embedding as an input; generating a question-answer dataset in which question text and answer text related to the high-resolution image are matched; concatenating the integrated vision embedding with a question embedding corresponding to the question text, so as to generate a second vision-language embedding; and carrying out second training involving training the LLM to output the answer text by using the second vision-language embedding as an input.
G16H 30/40 - TIC spécialement adaptées au maniement ou au traitement d’images médicales pour le traitement d’images médicales, p. ex. l’édition
G16H 30/20 - TIC spécialement adaptées au maniement ou au traitement d’images médicales pour le maniement d’images médicales, p. ex. DICOM, HL7 ou PACS
G16H 50/70 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour extraire des données médicales, p. ex. pour analyser les cas antérieurs d’autres patients
H04L 51/02 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel en utilisant des réactions automatiques ou la délégation par l’utilisateur, p. ex. des réponses automatiques ou des messages générés par un agent conversationnel
100.
PREDICTION SYSTEM AND CONTROL METHOD THEREOF, AND TRAINING METHOD FOR PREDICTION SYSTEM
LG MANAGEMENT DEVELOPMENT INSTITUTE CO., LTD. (République de Corée)
Inventeur(s)
Jang, Kyo Chul
Choi, Jae Mu
Jang, Seung Jun
Choi, Seo Young
Hong, Young Hoon
Abrégé
The present invention relates to a prediction system, a control method thereof, and a training method for the prediction system. More particularly, the present invention relates to a prediction system capable of predicting a valid customer company (or valid customer) in a business-to-business (B2B) and/or business-to-consumer (B2C) business situation, a control method thereof, and a training method for the prediction system.
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
G06Q 10/06 - Ressources, gestion de tâches, des ressources humaines ou de projetsPlanification d’entreprise ou d’organisationModélisation d’entreprise ou d’organisation
G06N 20/20 - Techniques d’ensemble en apprentissage automatique
G06Q 30/02 - MarketingEstimation ou détermination des prixCollecte de fonds