Machine learning

SYSTEM AND METHOD FOR MEMORY CREATION

Resumen de: EP4647971A1

The present disclosure generally relates to generating a video corresponding to a memory (e.g., an event or context) from media assets on a device. In some embodiments, the device receives user inputs requesting a video based on a natural language description of a memory. The device sends information of the natural language description to a first machine-learning (ML) model, and receives query tokens, which are used to find media items on the device that match the query tokens. The device sends information representing the found media items to another ML model that determines traits from the media items. These traits are sent to a third ML model to generate a story outline, and the video is generated by comparing the descriptions of shots in the story outline to visual embeddings of the found media assets to curate and arrange them into the video consistent with the story outline.

Training generative artificial intelligence models

Resumen de: GB2640912A

A method for training generative AI / machine learning (ML) models wherein the training includes iterative steps and reinforcement learning wherein a reinforcement learning reward value for one model is based on a likelihood value obtained by another model. The method includes a set of rules and iterative training steps to train the generative ML models. Each iterative training step assigns to each model a role which includes an actor model and a judge model. Each step then prompts the assigned actor model with an input to generate content that complies with a constitution. Each step also prompts the assigned judge model with content generated by the actor model and determines a likelihood of compliance that the content complies with the constitution. Each iterative training step also provides a reinforcement learning reward for training which is based on the likelihood of compliance determined by the judge model. The method may allow for the models to be trained using reinforcement learning via the reward and by switching the roles of the models. The models may be used in natural language processing tasks (e.g. reasoning, decision making etc).

METHOD, SYSTEM, AND COMPUTER PROGRAM PRODUCT FOR IMPROVING MACHINE LEARNING MODELS

Resumen de: US2025165874A1

Methods, systems, and computer program products are provided for improving machine learning models which include receiving a data set including data records; inputting the data set to a pre-trained first machine learning model to generate first embeddings; inputting the first embeddings to a second machine learning model to generate second embeddings in a user-specific embedding space; inputting the plurality of second embeddings to a third machine learning model to extract feature data associated with a feature; inputting an output from a machine learning system and the feature data to a fourth machine learning model to generate a relevance score for each entity; determining a subset of entities based on the relevance score; communicating a feedback request to a user; receiving feedback data from the user; and training at least one of the models based on the feedback data.

PREDICTING TROPOSPHERIC DUCTING EVENTS

Resumen de: EP4648457A1

Disclosed is a method comprising collecting input data (521) comprising at least weather forecast information for an area in which one or more cells (104B, 104C, 104D) are located; providing the input data (521) to a prediction algorithm (520), wherein the prediction algorithm (520) comprises: a machine learning model (500) trained to predict tropospheric ducting events impacting the one or more cells (104B, 104C, 104D), and a cell site database (511) indicating a location and one or more configuration parameters of the one or more cells (104B, 104C, 104D); and receiving, from the prediction algorithm (520), output data (522) indicating one or more predicted tropospheric ducting events expected to impact the one or more cells (104B, 104C, 104D) based on the input data (521).

AUTOMATED MACHINE LEARNING TO GENERATE RECOMMENDATIONS FOR WEBSITES OR APPLICATIONS

Resumen de: US2025342373A1

Implementations described herein relate to methods, systems, and computer-readable media for automated generation and use of a machine learning (ML) model to provide recommendations. In some implementations, a method includes receiving a recommendation specification that includes a content type and an outcome identifier, and determining model parameters for a ML model based on the recommendation specification. The method further includes generating a historical user feature matrix (FM), generating a historical content feature matrix (FM), and transforming the historical user FM and the historical content FM into a suitable format for the ML model. The method further includes obtaining a target dataset that includes historical results for the outcome identifier for a plurality of pairs of user identifiers and content items of the content type. The method further includes training the ML model using supervised learning to generate a ranked list of content items for each user identifier.

QUANTUM FEATURE MAPS

Resumen de: AU2024261237A1

In a general aspect, a quantum feature-map transforms an input dataset with original features to a preprocessed dataset with quantum-enhanced features. In some cases, pre-processing an input dataset for a machine learning model includes obtaining the input dataset comprising a plurality of data points; encoding the plurality of data points as parameters of a quantum logic circuit; and executing the quantum logic circuit on a quantum computing resource. Expectation values of a set of observables are obtained based on an output quantum state generated by executing the quantum logic circuit. The set of observables includes observables of first degree and observables of second degree. A pre-processed dataset is generated based on the expectation values and provided as an input to the machine learning model.

SYSTEMS AND METHODS FOR BINDING AT LEAST ONE UNIQUE SCHEMA-SPECIFIC IDENTIFIER TO A CATEGORY

Resumen de: WO2025231033A1

A method including receiving activity data related to a first activity utilizing an unbound schema-specific identifier; training a machine learning engine based on at least one input to obtain a trained machine learning engine that is trained to identify a category associated with the entity; where the at least one input includes: an entity data feature vector, a historical user activity data feature vector, and/or a historical user schema-specific identifier data feature vector; predicting via the trained machine learning engine, a category associated with the first activity; binding the unbound schema-specific identifier to the category to generate a category bound schema-specific identifier; receiving a request to perform a second activity using the bound schema-specific identifier; determining if a second entity associated with the request to perform the second activity is associated with the category; performing one of: approving or denying the request to perform the second activity.

INFORMATION PROCESSING DEVICE, INFORMATION PROCESSING METHOD, AND PROGRAM

Resumen de: US2025342281A1

The present disclosure relates to an information processing device, an information processing method, and a program capable of effectively detecting counterfeit data using a more versatile method.A contribution indicating how much each feature in a training dataset contributes to a predicted label output from a trained model is calculated, the training dataset including both a legitimate sample including only legitimate data and a counterfeit sample at least partially including counterfeit data. Then, clustering is executed to classify each sample of the training dataset into a plurality of clusters using unsupervised learning with the contribution as input, and feature variability between the clusters in the result of the clustering is compared to identify a cluster to which the counterfeit sample included in the training dataset belongs. The present technology can be applied to, for example, a machine learning system that generates a fraud detection model.

METHODS AND APPARATUS FOR DETECTION OF MALICIOUS DOCUMENTS USING MACHINE LEARNING

Resumen de: US2025342250A1

An apparatus for detecting malicious files includes a memory and a processor communicatively coupled to the memory. The processor receives multiple potentially malicious files. A first potentially malicious file has a first file format, and a second potentially malicious file has a second file format different than the first file format. The processor extracts a first set of strings from the first potentially malicious file, and extracts a second set of strings from the second potentially malicious file. First and second feature vectors are defined based on lengths of each string from the associated set of strings. The processor provides the first feature vector as an input to a machine learning model to produce a maliciousness classification of the first potentially malicious file, and provides the second feature vector as an input to the machine learning model to produce a maliciousness classification of the second potentially malicious file.

COPILOT IMPLEMENTATION: RESTRICTING OPERATION TO A DOMAIN OF COMPETENCE

Resumen de: US2025342171A1

Apparatus and methods are disclosed for implementing a copilot as a network of microservices including specialized large language models (LLMs) or other trained machine learning (ML) tools. The microservice network architecture supports flexible, customizable, or dynamically determinable dataflow from client input to corresponding output. Compared to much larger competing LLMs, comparable or superior performance is achieved for certain tasks, while significantly reducing computation time and hardware requirements, even to a single compute node with a single GPU. Examples incorporate a qualification microservice to test data, destined for a downstream microservice, for conformance with the copilot's competency. A knowledge graph of a corpus of documents is built, visualized, and pruned. The data is tested for conformance with the pruned graph representation, and non-conforming data is excluded from the dataflow. Variations and additional techniques are disclosed.

MANAGING A PLURALITY OF WIRELESS DEVICES THAT ARE OPERABLE TO CONNECT TO A COMMUNICATION NETWORK

Resumen de: US2025344079A1

A method for managing a plurality of wireless devices. The method includes obtaining a plurality of base Machine Learning (ML) models, wherein each base ML model is operable to provide an output on the basis of which at least one RAN operation performed by a wireless device may be configured. The method further includes transmitting characterising information for individual models of the plurality of base ML models and configuration information for the plurality of base ML models over the RAN. The method further includes receiving an indication of which one or more of the plurality of base ML models the wireless device will be using as an ensemble ML model in connection with a RAN operation performed by the wireless device, and setting a value of at least one configuration parameter associated with the RAN operation performed by the wireless device based on the received indication.

PREDICTING TROPOSPHERIC DUCTING EVENTS

Resumen de: US2025344080A1

Disclosed is a method comprising collecting input data comprising at least weather forecast information for an area in which one or more cells are located; providing the input data to a prediction algorithm, wherein the prediction algorithm comprises: a machine learning model trained to predict tropospheric ducting events impacting the one or more cells, and a cell site database indicating a location and one or more configuration parameters of the one or more cells; and receiving, from the prediction algorithm, output data indicating one or more predicted tropospheric ducting events expected to impact the one or more cells based on the input data.

ESTIMATING THE RISK OF MEMBERSHIP INFERENCE ATTACKS ON MACHINE LEARNING MODELS

Resumen de: US2025343816A1

In various examples there is a method of empirically measuring a level of security’ of a training pipeline. The training pipeline is configured to train machine learning models using confidential training data. The method comprises storing a representation of a joint distribution of false positive rate and false negative rate of membership inference attacks on a plurality of machine learning models trained using the training pipeline. The method uses the representation to compute a posterior distribution of the level of security’ from observations of the membership inference attack on the plurality’ of machine learning models trained using the training pipelines. A confidence interval of the level of security is computed from the posterior distribution and the confidence interval is stored.

SYSTEMS AND METHODS FOR GENERATING A LIFESTYLE-BASED DISEASE PREVENTION PLAN

Resumen de: US2025342936A1

A system for generating a lifestyle-based disease prevention plan, the system including a computing device configured to receive at least a user biomarker input, produce a user profile as a function of the at least a user biomarker input, and generate a lifestyle-based disease prevention plan as a function of the user profile including training a machine learning process with a lifestyle training data set where the lifestyle training data set further comprises lifestyle elements correlated to a plurality of outputs containing diseases prevented and producing the lifestyle-based disease prevention plan as a function of the user profile and machine learning process.

Prediction of Alzheimer's Disease

Resumen de: US2025342959A1

A method for diagnosing Alzheimer's Disease or determining susceptibility to Alzheimer's Disease includes steps of obtaining a blood sample from a target subject and extracting cell-free (cf) DNA from the blood sample as extracted cf DNA. The degree of methylation in one or a plurality of Alzheimer indicator genes in the extracted cf DNA is identified. Each Alzheimer indicator gene identified is an indicator of the presence of or risk of developing Alzheimer's Disease where the plurality of Alzheimer indicators genes have been identified by a machine learning technique or by logistic regression. The target subject is identified as being at risk for Alzheimer's Disease if the amount of methylation of one or more Alzheimer's indicator genes differs from the amount of methylation established in control subjects not having Alzheimer's Disease to a statistically significant degree.

PRODUCING AN AUGMENTED DATASET TO IMPROVE PERFORMANCE OF A MACHINE LEARNING MODEL

Resumen de: US2025342394A1

Producing an augmented dataset to improve performance of a machine learning model. A test series is created for a first type of data transformation. the test series defining a set of test values for at least one parameter characterizing the first type of data transformation. Test datasets are generated based on a source dataset, each of the test datasets corresponding to a respective test value of the set of test values for said at least one parameter characterizing the first type of data transformation. Each of the test datasets is input to the machine learning model to produce a corresponding model output. At least one score is determined for each test dataset based at least in part on the corresponding model output. Robustness metrics of the first type of data transformation are determined based on a function which maps said at least one score of each of the test datasets to said at least one parameter characterizing the first type of data transformation. A set of one or more data augmentations are determined to be applied to the source dataset based at least in part on said one or more robustness metrics of the first type of data transformation. An augmented dataset is generated based on the source dataset using the determined set of one or more data augmentations.

MACHINE-LEARNING-BASED TECHNIQUES FOR DETERMINING RESPONSE TEAM PREDICTIONS FOR INCIDENT ALERTS IN A COMPLEX PLATFORM

Resumen de: US2025342374A1

Various embodiments of the present invention provide methods, apparatuses, systems, computing devices, and/or the like that are configured accurately and programmatically train a responder prediction machine learning model for generating response team predictions based on the systematic collection of one or more responder prediction training corpuses comprising one or more alert related datasets in a responder prediction server system. For example, the responder prediction server system may extract one or more alert attributes for each of the one or more alert related datasets for training one or more responder prediction machine learning models and/or one or more prioritization machine learning models. The responder prediction machine learning model and prioritization machine learning models may process one or more alerts, in real-time, to generate one or more response team prediction objects for rendering in a response team suggestion interface.

SYSTEMS AND METHODS FOR USING MACHINE LEARNING TECHNIQUES TO PREDICT ITEM GROUP COMPOSITION

Resumen de: CN120266139A

Systems and methods for predicting group composition of items are disclosed. A system for predicting group composition of items may include a memory storing instructions and at least one processor configured to execute the instructions to perform operations including: receiving entity identification information and a timestamp associated with a transaction without receiving information distinguishing items associated with the transaction; determining a localized machine learning model based on the entity identification information, the localized machine learning model trained to predict a category of an item based on transaction information applied to all of the items associated with the transaction; and applying the localized machine learning model to a model input to generate a predicted category of items associated with the transaction, the model input including the received entity identification information and timestamps but not information distinguishing items associated with the transaction.

METHOD AND APPARATUS FOR BEAM MANAGEMENT BY USING ARTIFICIAL INTELLIGENCE AND/OR MACHINE LEARNING

Resumen de: EP4645709A1

Provided are a method and an apparatus for performing beam management in a wireless communication system. The method of a terminal may include triggering at least one beam failure recovery (BFR) for a cell that performs beam management using an artificial intelligence and/or machine learning model, deactivating the artificial intelligence and/or machine learning model based on the number of the at least one beam failure recovery triggered during a specific time duration, and transmitting deactivation information of the artificial intelligence and/or machine learning model to a base station. The method of the base station may include transmitting, to the terminal, configuration information related to the artificial intelligence and/or machine learning model, receiving, from the terminal, the deactivation information of the artificial intelligence and/or machine learning model, and, based on the received deactivation information, stopping beam generation related to the artificial intelligence and/or machine learning model.

DETERMINING CONFIDENCE TO START VALUES FOR POWER GENERATION SYSTEMS USING MACHINE LEARNING MODELS

Resumen de: CN120500694A

Systems and methods for applying a machine learning (ML) model to determine a startup confidence value for a generator are presented herein. The computing system may identify a first plurality of parameters of the first generator. The plurality of parameters may identify operation of the first generator. The computing system may apply the first plurality of parameters to the ML model to determine a first confidence value that identifies a first likelihood that the first generator is started at initiation. The computing system may provide an output based on a first confidence value of the first generator.

INFORMATION PROCESSING SYSTEM AND PROGRAM

Resumen de: EP4645322A1

Comprising at least one processor obtaining a combination of information identifying each of the raw materials received from the user and the amount of each of the raw materials, and obtaining a predicted value of a physical property of the property name to be predicted for a composition comprising each of the raw materials by inputting into a first machine learning model at least one of the chemical fingerprints, SMILES strings or chemical graph structure data or product name or substance name corresponding to each of the raw materials and the amount of each of the raw materials, or by inputting into a second machine learning model a set of values based on at least one of the chemical fingerprints, SMILES strings or chemical graph structure data or product name or substance name corresponding to each of the raw materials and the amount of each of the raw materials,wherein the first machine learning model is a model in which parameters are adjusted so that it can predict outputs from inputs by means of a learning data set that takes as inputs at least one of the chemical fingerprints, SMILES strings or chemical graph structure data or product names or substance names corresponding to each of the raw materials and the amount of each of the said raw materials, and as takes as outputs physical property values of the target property names to be predicted, and the second machine learning model is a model in which parameters are adjusted so that it can predict outputs from inputs b

CELL MANUFACTURING MANAGEMENT PLATFORM USING MACHINE LEARNING

Resumen de: US2025336521A1

A cell manufacturing management platform facilitates management of a cell manufacturing process. The cell manufacturing management platform tracks events associated with a cell manufacturing process and coordinates between disparate entities involved in the process. The cell manufacturing management platform utilizes machine learning techniques to generate inferences associated with event scheduling in a manner that optimizes an efficiency metric and reduces likelihood of exceptions occurring. Machine learning models may furthermore be used to generate various alerts or other actions associated with the process. A user interface enables different participating entities to track progress of the process and upcoming events.

MACHINE LEARNING MODELS OPERATING AT DIFFERENT FREQUENCIES FOR AUTONOMOUS VEHICLES

Resumen de: US2025335796A1

Systems and methods include machine learning models operating at different frequencies. An example method includes obtaining images at a threshold frequency from one or more image sensors positioned about a vehicle. Location information associated with objects classified in the images is determined based on the images. The images are analyzed via a first machine learning model at the threshold frequency. For a subset of the images, the first machine learning model uses output information from a second machine learning model, the second machine learning model being performed at less than the threshold frequency.

SYSTEMS AND METHODS FOR DETERMINING A USER SPECIFIC MISSION OPERATIONAL PERFORMANCE METRIC, USING MACHINE-LEARNING PROCESSES

Resumen de: US2025335326A1

Aspects relate to system and methods for determining a user specific mission operational performance, using machine-learning processes. An exemplary system includes a computing device configured to perform operations including receiving user-input structured data from at least a user device, receiving observed structured data related to the user and a mission performance metric, inputting the user-input structured data and the observed structured data to a machine-learning model, generating a user performance metric as a function of the machine-learning model, receiving a deterministic mission operational performance metric, disaggregating a deterministic user performance metric as a function of the deterministic mission operation performance metric and the mission performance metric, inputting training data to a machine-learning algorithm, where the training data includes the user-input structured data and the observed structured data correlated to the deterministic user performance metric, and training the machine-learning model as a function of the machine-learning algorithm and the training data.

AI-Based Energy Edge Platform, Systems, and Methods Having a Digital Twin of Decentralized Infrastructure

Nº publicación: US2025334943A1 30/10/2025

Solicitante:

STRONG FORCE EE PORTFOLIO 2022 LLC [US]
Strong Force EE Portfolio 2022, LLC

Resumen de: US2025334943A1

An AI-based platform for enabling intelligent orchestration and management of power and energy is provided herein. The AI-based platform includes a digital twin system including a plurality of digital twins of energy operating assets, the plurality of digital twins of energy operating assets including at least one energy generation digital twin, energy storage digital twin, energy delivery digital twin, and/or energy consumption digital twin, and a set of energy simulation systems configured to generate a simulation of energy-related behavior of at least one of the plurality of digital twins of energy operating assets, and a machine-learning system configured to generate a predicted state of at least one of the energy operating assets. The simulation of energy-related behavior is based on historical patterns, current states, and the predicted state of at least one of the energy operating assets.