Alerta

PREDICTION-GUIDED ENSEMBLING FOR MACHINE LEARNING MODELS

Resumen de: US20260105383A1

An approach is provided for prediction-guided machine learning model ensembling. Label groupings within an output range of a base machine learning model are determined. Accuracies of the base model across the label groupings are evaluated. One or more of the label groupings are identified whose respective evaluated accuracy does not satisfy end user-defined criteria. Using a reduced training set, a specialized machine learning model for a given label grouping included in the identified one or more label groupings is trained. A majority of samples of the reduced training set are from the given label grouping. During inference, it is determined that an initial prediction by the base model is within an output range specified by the given label grouping. A weighting for an ensembling using the base model and the specialized model is determined. Using the ensembling, the initial prediction is refined to generate a final prediction.

MODEL INFERENCE METHOD AND APPARATUS

Resumen de: US20260105166A1

A model inference method and apparatus are disclosed, and relates to the field of machine learning technologies. A client and a server use respective deployed models to process different parts of user data, to obtain respective output results. In addition, the client obtains the output result of the server, and obtains an inference result based on the output results of the server and the client. Compared with a case in which the server needs to obtain all the user data in an inference process, in this application, the server obtains only a part of the user data. As the server cannot obtain, based on the part of the user data, all content included in the user data, security of the user data is ensured.

SYSTEMS AND METHODS FOR AUTOMATED MACHINE LEARNING

Resumen de: US20260105354A1

Embodiments of the present disclosure include techniques for automatically generating machine learning models. In one embodiment, sets of hyperparameters corresponding to machine learning models trained on one training data set are provided as an input. The hyperparameters are iteratively selected using an algorithm, such as a bandit algorithm, and used to train an ML model using another training data set. The performance of the trained ML model is evaluated on each iteration until the ML model performance is above a threshold. The resulting model may be used to train a resulting model. In some embodiments, ML models are combined across iterations to improve performance.

PRODUCT DESCRIPTION GENERATION WITH MACHINE LEARNING

Resumen de: US20260105504A1

0000 A system can include one or more memory devices storing instructions thereon that, when executed by one or more processors, can cause the one or more processors to retrieve information for a product of an entity, determine that the information includes proprietary information associated with the product or the entity, modify the information to remove the proprietary information from the information, input the modified information into a machine learning model, and generate a second description of the product using the modified information.

WOOD WIDE MODELS FOR IMPROVED MACHINE LEARNING PIPELINE

Resumen de: WO2026080714A1

The invention is systems and methods directed to a method of training a machine learning model that is particularly adept at handling numeric data and is able to learn from prior models trained using the same method.

Machine Learning Model Provenance Based on Computational Graphs

Resumen de: US20260105361A1

The provenance of a machine learning model can determined by receiving at least one file encapsulating the machine learning model. A computational graph corresponding to the machine learning model is extracted from the at least on file. The computational graph is converted from a first format into a normalized computational graph having a second, different format. The normalized computational graph is decomposed into components. These components can include, for example, nodes, edges between nodes, blocks, and edges between blocks. The components of the normalized computational graph can be compared with components of each of a plurality of normalized computational graphs which, in turn, each correspond to a different known machine learning model. Based on the comparison, a model genealogy of the machine learning model is determined. Data characterizing this model genealogy is provided to a consuming application or process.

DYNAMIC TRAFFIC ALLOCATION USING MACHINE LEARNING

Resumen de: US20260105118A1

0000 Methods, systems, and apparatus, including computer programs encoded on computer storage media, for dynamically managing webpage traffic. In some implementations, a server receives a request from a client device to access a webpage. The server obtains attributes associated with the client device that provided the request. The server selects one or more webpage variants using the request. Data associated with the request and with the one or more webpage variants are provided as input to a trained machine learning model. The server obtains output from the machine learning model that indicates the likelihood for each webpage variant that the user will access a particular feature of the corresponding webpage variant. The server provides data representing the webpage variant of the one or more webpage variants whose likelihood satisfies a threshold value.

VALIDATING USE OF DATA IN TRAINING OF MACHINE LEARNING MODELS

Resumen de: US20260105327A1

Techniques for validating use of data in training of machine learning (ML) models are disclosed. Synthetic data is by generated, by sampling from a statistical distribution of user data. The user data and the synthetic data are fed to an inference endpoint of the ML model. First results are generated by the ML model, based on the user data; and second results are generated by the ML model, based on the synthetic data. A statistical analysis is conducted, based at least in part on the first results and the second results. A determination is made as to whether the user data was used for training the ML model, based at least in part on conducting the statistical analysis. An indication of the determination as to whether the user data was used for training the ML model is displayed on a user interface.

Method and Apparatus for Training and Employing a Machine Learning Model to Identify Failed Components

Resumen de: US20260105359A1

A trained machine learning model identifies that a real-world apparatus has a failed component, which trained machine learning model has been trained with a training corpus that includes content generated by synthesizing a plurality of synthesized operating examples for a given apparatus, wherein at least some of the plurality of synthesized operating examples are generated via a simulation modeling environment that receives as input characterizing information that corresponds to any of a variety of failure states for a component of the given apparatus.

SYSTEMS AND METHODS FOR GENERATING RESEARCH PREDICTIONS

Resumen de: US20260105374A1

The disclosed technology can include a system capable of receiving a corpus of documents including a first subset of documents and a second subset of documents, where the first subset of documents and the second subset of documents are received at different time intervals, generating a credibility score and an impact score for each document of the first subset of documents, selecting a training subset from the first subset of documents based on the credibility score and the impact score, training a machine learning algorithm based on the training subset, generating, using the machine learning algorithm, a plurality of hypotheses, and evaluating the plurality of hypotheses against the second subset of documents.

OPTIMIZED USE OF PRIVACY BUDGET

Resumen de: US20260105373A1

Method comprising determining, in a trusted execution environment, values of hyperparameters of a machine learning model based on private data stored in the trusted execution environment, wherein the hyperparameters include system-specific hyperparameters and model-specific hyperparameters; training, in the trusted execution environment, the machine learning model to which the determined values of the system-specific and model-specific hyperparameters are applied to obtain, after one or more epochs of training, a sufficiently trained machine learning model; outputting the sufficiently trained machine learning model from the trusted execution environment; and, inhibiting output of the determined values of the system-specific hyperparameters from the trusted execution environment, wherein the system-specific hyperparameters are not accessible in the outputted sufficiently trained machine learning model.

QUANTUM KERNEL ENHANCED DEEPFAKE DETECTION AND PREVENTION

Resumen de: US20260106877A1

System, method, and computer program product embodiments detect synthetic media known as deepfakes based on received inferencing data that includes content such as a file or stream of audio, image, video, or textual chat data. The inferencing data further includes inferencing metadata associated with a source of the content. A data vector derived from one or more samples of the content and the inferencing metadata is transmitted to quantum computing hardware for quantum amplitude encoding of the data vector into a set of qubits, which are processed with a trained quantum support vector machine (SVM) to produce a classification of the content as synthetic or genuine. Based on a signal indicating that the content is synthetic, a notification or alert is generated and sent to a human user or an automated system warning that the content is classified as including deepfake data.

TECHNIQUES FOR USING KNOWLEDGE GRAPHS TO AUTOMATICALLY COMPLETE DRAFT STRUCTURAL DESIGNS

Resumen de: US20260105219A1

0000 One embodiment sets forth a technique for completing computerized representations of physical structures using knowledge graphs. According to some embodiments, the technique includes the steps of generating a knowledge graph that characterizes relationships between various features of computerized representations of physical structures; training one or more machine learning models based on the knowledge graph; receiving a request for adding a selected feature to a computerized representation of a physical structure; generating predicted feature data using the one or more trained machine learning models and the request; and causing the predicted feature data to be rendered at a graphical user interface (GUI) to suggest a predicted feature to be included with the selected feature. Another embodiment sets forth a technique for training machine learning models using knowledge graphs associated with computerized representations of physical structures.

METHODS AND SYSTEMS FOR ASSESSMENT OF FORECASTABILITY OF A TIME SERIES

Resumen de: US20260105376A1

0000 Systems and methods are disclosed for assessing the forecastability of time series data using machine learning. A large set of time series is preprocessed and partitioned into training and testing sets. Multiple forecasting models are trained on the training data, and the best-performing model is used to compute a sample forecastability metric. Features are extracted from the time series and paired with the forecastability metric to train a predictive model. This trained model is then used to estimate the forecastability of new time series data based on extracted features, bypassing the need for computationally intensive forecasting. The approach enables efficient and reliable forecastability assessment across large volumes of time series data.

ARTIFICIAL INTELLIGENCE CHATBOT DATA PARSER

Resumen de: WO2026080665A1

Some non-limiting aspects of the present disclosure describes generating dataset for implementing a rules-driven query system on a machine learning model. With this method, the system can interchange data easily since little modifications would be needed to shift this approach to using any other information. In this way, the present disclosure describes reducing bottlenecks for sales personnel and technicians to access parts data to complete quotations and service.

HEALTH TRACKING APPLICATIONS FOR SMART GLASSES

Resumen de: WO2026080545A1

Systems, computer programs, devices, and methods that enable coordination across multiple devices of the mobile ecosystem. In one embodiment, smart glasses detect when a user is about to eat food or take a drink and capture the consumable and portion. The data is recorded in a "morsel track" for health activity analysis. Low-fidelity captures provide preliminary recognition, while higher-fidelity captures are selectively invoked for definitive classification. Machine-learning logic generates predicted metabolic responses, such as real-time glucose trends, based on the recorded events. Predicted responses may dynamically adjust the operation of continuous glucose monitors, heart-rate sensors, or other biomedical devices. In some embodiments, the system triggers a pharmaceutical dispenser, such as an insulin pump, inhaler, or transdermal patch, to provide closed-loop therapeutic intervention in real time.

METHOD AND SYSTEM FOR RISK CONTROL MODELLING AND DEPLOYMENT

Resumen de: WO2026080013A1

In a described embodiment, a system for predictive modeling is provided. The system includes a data input module configured to acquire raw data and a data pre-processing module 5 configured to process the raw data to generate sample data. The system further includes a feature engineering module configured to generate derived features using one or more specified criterion and a model generation and optimization module configured to generate and refine a predictive model using a plurality of machine learning techniques. A model evaluation module is configured to assess model performance using predefined metrics. A 10 model tuning system is configured to iteratively adjust model parameters of the predictive model based on performance feedback and a model deployment module is configured to implement the predictive model in a production environment.

SYSTEMS AND METHODS FOR AUTOMATED CONFIGURATION TO ORDER AND QUOTE TO ORDER

Resumen de: AU2026202443A1

Computerized systems and methods are disclosed for automating Configure to Order (CTO) and Quote to Order (QTO) processes. Methods include receiving user inputs for desired product configurations, retrieving corresponding data from a bill of materials database, and calculating optimized pricing through intelligent rules based on real-time market data. Automated quotes are generated and transferred to orders in a vendor system, selected based on pre-set criteria like vendor reputation and delivery time. Validation steps reduce errors, and real-time reports are generated. The system integrates a Real-Time Data Mesh for data aggregation, a Single Pane of Glass User Interface for user interactions, and Advanced Analytics and Machine Learning Modules for implementing rule-based and learning algorithms. The system is accessible across various devices and standardizes data for uniform consumption, while also employing machine learning models to continually optimize processes. Notifications are sent to users upon successful execution of orders or completion of quotes. ar a r

CHATBOTS COMBINING ARTIFICIAL INTELLIGENCE AND DATABASE PROCESSING

Resumen de: US20260106847A1

Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for combining artificial intelligence and database processing in chatbots and other applications and interfaces. In some implementations, a system receives a user prompt through the interface. In response, the system obtains code or instructions generated using one or more artificial intelligence or machine learning (AI/ML) models, where the code or instructions are configured to cause a data processing system to retrieve data relevant to the user prompt from a data set. The system obtains result data as a result of processing the code or instructions using a data processing system. The system obtains natural language text generated by the one or more AI/ML models using the result data, and the system provides a response to the user prompt that is based on the natural language text generated by the one or more AI/ML models using the result data.

PERFORMANCE TEST FOR FUNCTIONALITY CHANGE

Resumen de: WO2026078301A1

Example embodiments of the present disclosure provide a solution for a performance test caused by a functionality change. In an example method, a terminal device determines a change in an artificial intelligence / machine learning model of a functionality of the terminal device that is connected to a radio access network. Then, the terminal device transmits a functionality applicability report for the model of the functionality, wherein the functionality applicability report includes a reason of a change in the functionality, an applicability indication for the functionality and model status of the AI/ML model. Next, the terminal device receives a configuration message indicating at least one test configuration for a performance testing procedure for the AI/ML model of the functionality, wherein the at least one test configuration includes at least one procedure parameter for the performance testing procedure determined based on the reason, the applicability indication and the model status.

METHOD AND DEVICE FOR CONFIGURING INFERENCE RESULT REPORTING UNIT IN MACHINE LEARNING-BASED BEAM MANAGEMENT

Resumen de: WO2026079733A1

This method of a terminal may comprise the steps of: identifying, from inference result reporting for a plurality of time instances, a candidate set related to the number of beams to be included in a reporting unit for differential reporting; receiving, from a base station, information indicating a first number belonging to the candidate set; and transmitting, to the base station, an inference result report including the reporting unit that includes inference results for the number of beams corresponding to the first number.

METHOD AND APPARATUS FOR TRANSMITTING UCI FOR REPORTING INFERENCE RESULT IN MACHINE LEARNING-BASED BEAM MANAGEMENT

Resumen de: WO2026079735A1

A method for a user equipment comprises the steps of: receiving, from a base station, configuration information about the number Nt (where Nt is a natural number of 1 or more) of a plurality of time instances; receiving, from the base station, configuration information about the number Nb (where Nb is a natural number of 1 or more) of beams to be reported for each of the plurality of time instances; generating an inference result report comprising inference results for Nt*Nb beams; and transmitting the inference result report to the base station, wherein the inference result report comprises one or more reporting units, each of the one or more reporting units comprises inference results for K (where K is a natural number of 1 or more) or less beams, the number of the one or more reporting units is M*Nt, and M may be defined as M=ceiling (Nb/K).

SYSTEM AND METHOD FOR CONTEXTUAL DISCOVERY AND PRIORITIZATION OF HARDWARE PROCESSORS FOR EXECUTION OF ARTIFICIAL INTELLIGENCE TOOL MACHINE LEARNING MODEL ALGORITHMS ON AN INFORMATION HANDLING SYSTEM

Resumen de: US20260105357A1

0000 An information handling system includes a hardware processor with the hardware processor executing an AI productivity tool software module to invoke a plurality of ML model algorithms to identify a responsive capability intent action based on received user-query input, a system environment component discovery software application to gather runtime telemetry data describing a current consumption state of a plurality of available in-band, side-band, and networked ML model algorithm execution provider hardware processors, and a workload orchestrator to receive the runtime telemetry data and determine when the workload orchestrator switches from a first ML model algorithm execution provider hardware processor used to execute at least one of the plurality of ML model algorithms to a second ML model algorithm execution provider hardware processor having less active processing and that is capable. Further, the workload orchestrator may determine when to switch size-variants of an ML model algorithm based on output confidence scores.

COMPUTING TECHNOLOGIES FOR REAL-TIME HYPERPARAMETER TUNING OF MACHINE LEARNING TRAINING PROCESSES VIA LANGUAGE MODELS

Resumen de: US20260105319A1

0000 This disclosure solves various technological problems described above by using language models (LMs) (e.g., large, small) to enable an autonomous adjustment algorithm that performs hyperparameter optimization within an active training session, with minimal or no human oversight. Resultantly, these improvements improve computer functionality by enabling more efficient hyperparameter searching by improving model performance, efficiency, scalability, time-to-market, and cost savings.

MACHINE LEARNING TRAINING FOR CHARACTERIZING WATER INJECTION AND SEISMIC PREDICTION

Nº publicación: US20260105364A1 16/04/2026

Solicitante:

SCHLUMBERGER TECH CORPORATION [US]

Resumen de: US20260105364A1

A method including receiving a reservoir model of a target under-ground region. The method also includes extracting, from the reservoir model, a historic pressure distribution in grid cells of the target underground region. The method also includes extracting, from the reservoir model, distances. Each distance represents a distance between a grid cell and a corresponding lineament in the target underground region. The method also includes receiving historic earthquake data of past earthquakes in the target underground region. The method also includes generating a vector. The vector includes features and corresponding values for at least i) the historic pressure distribution, ii) the distances, and iii) the historic earthquake data. The method also includes training a trained machine learning algorithm by recursively executing a machine learning algorithm on the vector until convergence.