Machine learning

METHOD FOR DETERMINING SPARSE INTERACTION EFFECT OF BLACK-BOX ARTIFICIAL INTELLIGENCE MODEL

Absstract of: EP4700652A1

The present application relates to the technical field of machine learning. Disclosed are a method and system for interpreting a sparse interaction effect modeled by a black-box artificial intelligence model. The method and system can automatically analyze an interactive distribution modeled by a model. The implementation of the method and system comprises the following steps: providing data that needs to be assessed; using a black-box model to perform prediction on the data, so as to obtain a prediction result of the model; on the basis of an output of the black-box model, modeling the interaction effect between input units of samples, calculating the interaction intensity between combinations formed by the input units, and expressing the black-box model as an "AND addition relationships" and an "OR addition relationships" between the combinations of the input units; and performing optimization, such that the "AND addition relationships" and the "OR addition relationships" are sparser. The advantages of the present invention lie in that a quantification method for interpreting the interaction modeled by a black-box artificial intelligence model is provided, and in comparison with previous research, a sparser and concise interactive interpretation can be obtained.

BATCH SELECTION POLICIES FOR TRAINING MACHINE LEARNING MODELS USING ACTIVE LEARNING

Absstract of: US2024354655A1

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a machine learning model. In one aspect, a method comprises: generating a set of candidate batches of model inputs; generating, for each candidate batch of model inputs, a respective score for the candidate batch of model inputs that characterizes: (i) an uncertainty of the machine learning model in generating predicted labels for the model inputs in the candidate batch of model inputs, and (ii) a diversity of the model inputs in the candidate batch of model inputs; and selecting the current batch of model inputs from the set of candidate batches of model inputs based on the scores; and training the machine learning model on at least the current batch of model inputs.

MACHINE LEARNING (ML)-BASED METHOD FOR DETERMINING A CONTROL CHANNEL ELEMENT (CCE) AGGREGATION LEVEL FOR A USER EQUIPMENT (UE) IN A PHYSICAL DOWNLINK CONTROL CHANNEL (PDCCH)

Absstract of: WO2024218535A1

The disclosure relates to a ML-based method for determining a CCE aggregation level for a UE in a PDCCH. The method comprises obtaining RBS traces. The method comprises training, using first data obtained from the traces, a machine learning model to predict a probability of discontinuous transmission (DTX) "isDTX probability". The method comprises inputting second data obtained from the traces into the machine learning model, obtaining the isDTX probability and expanding the second data with the isDTX probability. The method comprises, for each of a plurality of probability thresholds (PTs) and for each of a plurality of strategies, selecting a data having an isDTX probability greater or equal to the PT and best satisfying the strategy and using the data to train a classifier. The method comprises selecting one classifier and using the classifier for determining the CCE aggregation level for the UE in the PDCCH.

SYSTEM FOR PREPARING MACHINE LEARNING TRAINING DATA FOR USE IN EVALUATION OF TERM DEFINITION QUALITY

Absstract of: EP4700604A1

A system for preparing machine learning training data for use in evaluation of term definition quality. The system can include a server having at least one server processor and at least one server memory for storing a plurality of terms with corresponding definitions, and a plurality of client devices each having at least one client memory device and at least one client processor. The client processor programmed to receive at least one of the plurality of terms and its corresponding definition from the server, display the term and its corresponding definition, and receive an indication of whether the definition satisfies one or more definition quality guidelines. The server memory includes instructions for causing the at least one server processor to receive the indications from the plurality of client devices and label each definition as satisfying each of the definition quality guidelines or not based on the received indications.

SYSTEMS AND METHODS IMPLEMENTING AN INTELLIGENT OPTIMIZATION PLATFORM

Absstract of: EP4700664A2

A system and method includes receiving a tuning work request for tuning an external machine learning model; implementing a plurality of distinct queue worker machines that perform various tuning operations based on the tuning work data of the tuning work request; implementing a plurality of distinct tuning sources that generate values for each of the one or more hyperparameters of the tuning work request; selecting, by one or more queue worker machines of the plurality of distinct queue worker machines, one or more tuning sources of the plurality of distinct tuning sources for tuning the one or more hyperparameters; and using the selected one or more tuning sources to generate one or more suggestions for the one or more hyperparameters, the one or more suggestions comprising values for the one or more hyperparameters of the tuning work request.

USER PROFILING USING CHAIN-OF-THOUGHT KNOWLEDGE GRAPHS FOR QUERYING A MACHINE LEARNING SYSTEM

Absstract of: EP4700603A1

Techniques are disclosed for a machine learning model, such as a large learning model (LLM), that incorporates a model of a chain of thought of a particular user when responding to a query from the user. In one example, a system generates a knowledge graph of a chain of thought of the user. The knowledge graph comprises nodes representing topics present within past queries by the user and edges representing a co-occurrence between the topics. The system determines, based on a topic present within a query from the user and the knowledge graph, a goal query comprising a goal topic. The system provides, to a machine learning model, the user to generate, by the machine learning model, a response. The machine learning model is constrained to include the goal topic of the goal query within the response. The system outputs, for display, the response to the query.

BALANCED TRAINING DATASETS FOR PREDICTING AIRCRAFT COMPONENT FAULTS

Absstract of: EP4700611A1

The present disclosure provides a method of generating a balanced training dataset for a machine learning model in one aspect, the method including: receiving flight sensor data corresponding to a plurality of flights, and applying one or more criteria to the flight sensor data to generate a training dataset including a plurality of first instances corresponding to flights of the plurality of flights. The method further includes assigning, using component fault data, respective labels to the plurality of first instances, and generating, for groups of one or more labels of the respective labels, a respective plurality of flight series. Each flight series includes a respective sequence of second instances that is based on some of the plurality of first instances, and that concludes with a second instance that is assigned a label included in the group.

METHODS AND SYSTEMS FOR REAL-TIME RESOLUTION OF ERRORS THAT ARE DETECTED BY MULTIVARIATE MACHINE LEARNING MODELS BETWEEN USERS USING UNIVARIATE RESPONSES

Absstract of: US20260050503A1

Methods and systems are for generating real-time resolutions of errors arising from user submissions, computer processing tasks, etc. For example, the methods and systems described herein recite improvements for detecting errors in one or more user submissions and providing resolutions in real-time. To provide these improvements, the methods and systems use a machine learning model that is trained to return probability error scores based on a plurality of variables. By using the multivariate approach, the methods and systems may produce a highly accurate detection.

APPARATUS AND METHOD OF TRANSPORT DATA AGGREGATION

Absstract of: US20260049833A1

An apparatus and method for transport management is presented. The apparatus includes a memory communicatively connected to a processor to output routing data of transport entities as a function of aggregated transport data, wherein the outputting comprises: receive transport data and bound parameters of a transport from a carrier device; iteratively train an aggregation machine-learning model to combine the transport data, wherein the training comprises generating an aggregation training data correlating the transport data as inputs and aggregated transport data as outputs; modify a characteristic of the transport; update the aggregated transport data based on the modification of the characteristic of the transport; retrain the aggregation machine-learning model as a function of the updated aggregated transport data; generate the routing data, wherein the routing data comprises instructions to further modify the characteristic of the transport; and automatically change the characteristic of the transport based on the routing data.

SHIP RECOURSE OPTIMIZATION

Absstract of: US20260043656A1

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining elements of a shipping network. One of the methods includes obtaining environmental input data, wherein the environmental input data includes weather forecast data; providing the environmental input data to a circulation model; and providing output environmental condition from the circulation model to a machine learning model trained to generate a route for a ship.

DEVICES, METHODS, APPARATUSES, AND COMPUTER READABLE MEDIA FOR FALLBACK OF MACHINE LEARNING FUNCTIONALITY

Absstract of: WO2026032684A1

Disclosed are devices, methods, apparatuses, and computer readable media for fallback of machine learning functionality An example apparatus for a terminal device may include at least one processor and at least one memory. The at least one memory may store instructions that, when executed by the at least one processor, may cause the apparatus at least to: receive from a network, at least one first configuration for a machine learning functionality of a determined network function, and a second configuration for a non-machine learning functionality of the determined network function, wherein the second configuration is a fallback configuration from the first configuration; receive from the network, a first indication indicating the terminal device to activate fallback from the machine learning functionality; and in response to the first indication, apply modifications to the first configuration for use during fallback, and enable the second configuration in the network function.

METHOD AND DEVICE FOR SELECTING BEAM TO BE REPORTED IN MACHINE LEARNING-BASED BEAM MANAGEMENT

Absstract of: WO2026034877A1

The present invention relates to a method by which a terminal selects a beam to be reported in machine learning-based beam management, the method comprising the steps of: receiving, from a base station, configuration information of a measurement resource set and M number of report beams for AI/ML inference; determining, on the basis of measurement values of the measured beams, a beam to be reported; and transmitting the determined beam information to the base station, wherein, when the number of candidate beams to be reported exceeds M due to tie beams having the same or similar measurement values, the final beams to be reported are determined by excluding at least one from among same through a tie beam processing operation.

MACHINE LEARNING FOR VIDEO GAME HELP SESSIONS

Absstract of: US20260042011A1

The disclosed concepts relate to training a machine learning model to provide help sessions during a video game. For instance, prior video game data from help sessions provided by human users can be filtered to obtain training data. Then, a machine learning model can be trained using approaches such as imitation learning, reinforcement learning, and/or tuning of a generative model to perform help sessions. Then, the trained machine learning model can be employed at inference time to provide help sessions to video game players.

TECHNIQUES FOR PROVIDING INTERACTIVE CLINICAL DECISION SUPPORT FOR DRUG DOSAGE REDUCTION

Absstract of: US20260045348A1

Examples described herein generally relate to recommending drug dosage reductions for a patient. A computer system may generate an initial non-linear glide path of recommended dosages starting at an initial dosage of a drug for a patient and ending at a goal dosage at an estimated time of arrival. The system may receive periodic patient monitoring including at least one drug withdrawal scale score, anxiety scale score, and indicated side effect. The system may determine, using one or more machine learning algorithms, a revised glide path based on a data record for the patient, the at least the drug withdrawal scale score and the at least one anxiety scale score for the patient. The system may recommend at least one medication or therapy for the indicated side effect. The system may determine a prescription adjustment based on the revised glide path.

METHOD AND SYSTEM FOR USING ROBOTIC PROCESS AUTOMATION TO PROVIDE REAL-TIME CASE ASSISTANCE TO CLIENT SUPPORT PROFESSIONALS

Absstract of: US20260044798A1

A case assistant is provided to client support professionals, which utilizes robotic process automation (RPA) technologies to analyze large amounts of data related to historical client cases that are similar to current open cases, data related to skilled experts associated with similar client cases, and data related to business exceptions. Several processes are utilized to provide this data to client support professionals, including a document similarity finder that utilizes a vector data collector, a tokenizer, a stop word remover, a relevance finder, and a similarity finder, several of which utilize a variety of machine learning technologies. Additional processes include a skilled experts finder and a business exceptions finder.

CONSISTENCY VERIFICATION OF AIML POSITIONING CONFIGURATIONS

Absstract of: WO2026033326A1

An apparatus including at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: transmit, to a network entity, a configuration used when at least one model is trained; wherein the at least one model is an artificial intelligence or machine learning model; and receive, from the network entity, information related to a consistency between the configuration used when the at least one model is trained and a configuration used when the at least one model is to be applied during inference.

METHODS AND SYSTEMS FOR DATA COLLECTION FOR ARTIFICIAL INTELLIGENCE BASED POSITIONING MODEL TRAINING IN A WIRELESS NETWORK

Absstract of: WO2026035512A1

A network device (PRU, WTRU) may receive a request to collect data for artificial intelligence or machine learning (AI/ML) positioning model training, for example from a network data analytics function (NWDAF) and/or from a model training logical function (MTLF) (450b). The request may include an indication of an area of interest, a time window associated with the data for AI/ML positioning model training, a requested number of data samples of the data for AI/ML positioning model training, and/or a data source type of the data for AI/ML positioning model training. The network device may receive the data for AI/ML positioning model training and/or receive location data associated with the one or more WTRUs. The network device may send the location data and the data for AI/ML positioning model training to the NWDAF or the MTLF (485, 495).

WINDOW CONFIGURATION FOR A REFERENCE SIGNAL RECEIVE RESOURCE-BASED PROCESSING TASK ASSOCIATED WITH AN ARTIFICIAL INTELLIGENCE MACHINE LEARNING MODEL

Absstract of: WO2026035375A1

Aspects of the disclosure are directed to a (e.g., capability-based window) configuration for a reference signal receive (RS-Rx) resource-based processing task associated with an artificial intelligence machine learning (AIML) model. In an aspect, the RS-Rx resource-based processing task may be related to sensing or positioning or another task type (e.g., beam management, channel state information (CSI) operations, etc.). In an aspect, the RS-Rx task may be associated with any type of RS-Rx resource relative to the UE (e.g., downlink reference signals, sidelink reference signals, etc.). Such aspects may provide various technical advantages, such as AIML processing window configurations that are configured based on AIML model-specific capabilit(ies) of the UE, which may improve functionalities associated with the AIML model (e.g., improved sensing or positioning or beam management, etc.) and/or improved AIML model monitoring.

AUTOMATED TRANSLATIONS FOR AUTONOMOUS CHAT AGENTS

Absstract of: US20260044690A1

Disclosed are various embodiments for automated translations for autonomous chat agents. A build service can send a translation request to a machine translation service, the translation request comprising training data in a first language and the translation request specifying a second language. The build service can then receive translated training data from the machine translation service, the translated training data having been translated from the training data into the second language. Next, the build service can create a translated workflow that comprises a translated machine learning model and a translated intent. Subsequently, the build service can add the translated training data to the translated workflow and train the translated machine learning model using the translated training data.

MEMORY-CONSTRAINED ATTENTION IN MACHINE LEARNING MODELS

Absstract of: US20260044745A1

Certain aspects of the present disclosure provide techniques and apparatus for machine learning. In an example method, a machine learning model comprising a plurality of layers, and a set of input data for the machine learning model, are accessed. A combination of hyperparameters for the machine learning model is selected based on the set of input data, comprising selecting, for each respective layer of the plurality of layers, a respective cache size based on the input data. The machine learning model is deployed according to the combination of hyperparameters.

MEMORY-CONSTRAINED ATTENTION IN MACHINE LEARNING

Absstract of: WO2026035335A1

Certain aspects of the present disclosure provide techniques and apparatus for machine learning. In an example method, a machine learning model comprising a plurality of layers, and a set of input data for the machine learning model, are accessed. A combination of hyperparameters for the machine learning model is selected based on the set of input data, comprising selecting, for each respective layer of the plurality of layers, a respective cache size based on the input data. The machine learning model is deployed according to the combination of hyperparameters.

MULTIVARIABLE SERVICE TERMINATION RISK CLASSIFICATION USING MACHINE LEARNING

Absstract of: US20260044803A1

A method can include receiving input data comprising a plurality of features for a plurality of users. A method can including providing the input data to a risk prediction model configured to predict a termination likelihood for each user. In some implementations, the risk prediction model can be a random forest model. A method can include identifying, based on the predicted termination likelihood for each user, an at risk population including users with a termination risk above a threshold amount. A method can include determining, for each user of the at risk population, a profile type of a plurality of profile types. The profile type can describe certain attributes of the user. In some implementations, an end user can select a profile type. A method can include outputting members of the at risk population having the selected profile type.

MACHINE LEARNING FOR VIDEO GAME HELP SESSIONS

Absstract of: WO2026035326A1

The disclosed concepts relate to training a machine learning model to provide help sessions during a video game. For instance, prior video game data from help sessions provided by human users can be filtered to obtain training data. Then, a machine learning model can be trained using approaches such as imitation learning, reinforcement learning, and/or tuning of a generative model to perform help sessions. Then, the trained machine learning model can be employed at inference time to provide help sessions to video game players.

BIOMASS USE ASSISTANCE DEVICE, METHOD, AND PROGRAM

Absstract of: EP4693123A1

A biomass utilization support device: acquires biomass information relating to a biobased material and product information for each of a plurality of products including information about materials configuring the products; uses a machine learning model, which has been trained to estimate appropriate values for replacement amounts in a case of replacing a portion of the materials configuring the products with the biobased material, and the acquired biomass information and product information to estimate the appropriate values for each of the plurality of products; calculates, for each of the plurality of products, environmental impact indicators in a case in which a portion of the materials configuring the products has been replaced with the biobased material at the replacement amounts represented by the estimated appropriate values; and outputs support information listing the estimated appropriate values and the calculated environmental impact indicators.

MACHINE LEARNING MODEL FEATURE SELECTION IN A COMMUNICATION NETWORK

Nº publicación: EP4690715A1 11/02/2026

Applicant:

ERICSSON TELEFON AB L M [SE]
Telefonaktiebolaget LM Ericsson (publ)

Absstract of: CN120898407A

Embodiments of the present disclosure provide machine learning model feature selection in a communication network. The method includes, in response to a feature selection trigger of a first machine learning model, determining a target input feature set for an analysis task based on contextual information related to the analysis task, the first machine learning model being currently provisioned for performing the analysis task based on a current input feature set, the current input feature set is different from the target input feature set; and causing a second machine learning model to be provisioned to perform an analysis task based on the determined set of target input features. In this manner, the machine learning model may be supplied with an optimized set of input features that is applicable to the current network context and provides an acceptable level of model performance.