Alerta

IDENTIFYING NOISE IN VERBAL FEEDBACK USING ARTIFICIAL TEXT FROM NON-TEXTUAL PARAMETERS AND TRANSFER LEARNING

Absstract of: US20260094061A1

Methods and systems are provided for classifying free-text content using machine learning. Free-text content (e.g., customer feedback) and parameter values organized according to a schema are received. A free-text corpus is generated, and an artificial-text corpus is generated by applying rules to the parameter values. The artificial-text corpus is generated by converting the parameter values into a finite set of words based on the rules and concatenating the words of the finite set of words into a fixed sequence wordlist. Feature vectors (e.g., sentence embeddings) based on the free-text corpus and the artificial-text corpus are combined and forwarded to a machine learning model for classification. The machine learning model may be trained with a bias towards a specified metric (e.g., precision, recall, F1 score). The model may be trained using transfer learning with training data from a different category of free-text content (e.g., a different category of customer feedback).

PREDICTIVE MACHINE LEARNING MODELS BASED ON DATA COLLECTED FROM A VARIETY OF SOURCES

Absstract of: WO2026072162A1

The computer-based methods and systems presented in this disclosure provide prediction occurrence of an event for an individual on a user device of the individual. The system receives, from a plurality of remote devices, pieces of input data about the individual. The system pre-processes the pieces of input data to make them ready to be processed by respective input modules of a machine learning model running on the system. Each input module is associated with a respective marker and processes the pre-processed data for that marker. Outputs of the input modules are further processed by the model. The model provides an output indicating respective probabilities that particular events happen. The system can generate one or more alerts based on the output of the model, and can send the alerts to contacts of the individual.

PRIVACY PRESERVING IDENTITY RESOLUTION USING DEEP LEARNING

Absstract of: WO2026072270A1

A system and method for privacy-preserving identity resolution using deep learning enables accurate matching of personally identifiable information (PH) while maintaining data security. The system employs a deep learning model trained with transformer architecture and contrastive learning on third-party identity graph data. Custom tokenizers process data by leveraging hierarchical structures and domain-specific characteristics. The trained model generates vector embeddings that enable fuzzy matching, accounting for variations in spellings, typographical errors, and data inconsistencies. A vector database stores embeddings for nearest neighbor searches to identify potential identity matches. The system enables identity resolution without requiring Pll data movement from first-party environments. The invention facilitates building accurate first-party identity graphs and enables secure collaboration between parties without exposing underlying Pll data.

HIERARCHICAL MONITORING OF ARTIFICIAL INTELLIGENCE OR MACHINE LEARNING MODELS FOR AIR INTERFACE

Absstract of: WO2026072207A1

Certain aspects of the present disclosure provide techniques for performing wireless communication. In some aspects, the techniques include obtaining a first metric associated with a model that is associated with wireless communication; and obtaining, in response to a trigger condition associated with the first metric being satisfied, a second metric associated with the model, the second metric providing a different measure of the model than the first metric.

SYSTEMS AND METHODS FOR FACILITATING CONVERSATION BASED ON TEXT DATA AND SPEECH DATA

Absstract of: AU2024407921A1

A method includes receiving a user input and generating a set of user input tokens based on the user input. The method also includes generating a set of enhanced input tokens by providing the set of user input tokens as input to a first machine learning model. A state is determined based on a previous state and at least one of the set of user input tokens or the set of enhanced input tokens. Predetermined data is retrieved from a database based on the state and at least one of the set of user input tokens or the set of enhanced input tokens. The method also includes generating a set of response tokens by providing the set of user input tokens and the predetermined data as input to a second machine learning model. Based on the set of response tokens, a response is sent to a user device.

METHODS OF PREDICTING PROPERTIES OF A CHEMICAL SYSTEM USING SURROGATE MODELS

Absstract of: US20260094677A1

Methods of predicting physicochemical properties of a chemical system using a family of surrogate or reduced order models, trained on first principle simulation results. The models are created using machine learning techniques. The chemical system can be a complex multicomponent and multiphase system such as produced water.

Multi-Stage Federated Learning in Wireless Networks

Absstract of: US20260094032A1

first group of AI agents to train and report, per each AI agent of the first group, a respective first partial AI or machine learning (ML) (AI/ML) model to the AI manager, receive the first partial model from each AI agent of the first group, generate a first version of a global model from the first partial models, if the first version of the global model is determined to be trustworthy, select a second group of AI agents to train and report, per each AI agent of the second group, a respective second partial AI/ML model to the AI manager, receive the second partial models and aggregate the second partial models and the first version of the global model into a second version of the global model.

ENHANCED SOCIAL MEDIA PLATFORMS BASED ON MACHINE LEARNING BASED AGGREGATION OF TRENDING EVENTS

Absstract of: AU2024354389A1

An example method for automatic generation of content based on an aggregation of trending events is provided. The method includes determining, by a computing device, a topic of interest. The method also includes determining additional information related to the topic of interest. The determining of the additional information includes, generating a prompt based on the topic of interest, submitting the prompt to an information search and retrieval system, and retrieving the additional information as an output of the information search and retrieval system. The method also includes generating, by a generative artificial intelligence model, a piece of annotated content associated with the topic of interest. The piece of annotated content comprises media content annotated with at least one selectable graphical object that links to the additional information. The method also includes providing, by the computing device, the piece of annotated content.

Granular Taxonomy for Customer Support Augmented with AI

Absstract of: US20260094166A1

A computer-implemented method for augmenting customer support is disclosed in which a granular taxonomy is formed to classify tickets based on customer issue topic. A dashboard and user interface of performance metrics may be generated for the topics in the taxonomy.Recommendations may also be generated to aid servicing customer support issues for topics in the taxonomy. This may include generating information to aid in determining topics for generating automated responses or generating recommended answers for particular topics. In some implementations, an archive of historic tickets is used to generate training data for a machine learning model to classify tickets.

A METHOD FOR DETERMINING A HARDWARE ARCHITECTURE FOR A MACHINE LEARNING MODEL

Absstract of: EP4718234A1

The invention relates to a method (100) for determining an hardware architecture (3) for a machine learning model (50), comprising:- Providing (101) an initial hardware architecture (1), the initial hardware architecture (1) describing hardware components (2) and computing characteristics of said hardware components (2),- Providing (102) the machine learning model (50),- Converting (103) the machine learning model (50) to an intermediate representation, the intermediate representation depicting a topology and/or a temporal structure of the machine learning model (50) as a graph structure,- Analysing (104) the intermediate representation to determine a memory footprint of the machine learning model (50),- Determining (105) the hardware architecture (3) for the machine learning model (50) based on the initial hardware architecture (1) and a result of the analysing (104).Furthermore, the invention relates to a computer program, an apparatus, and a storage medium for this purpose.

AUTOMATED MACHINE LEARNING BASED WORKFLOW FOR TIMESERIES FORECASTING

Absstract of: EP4718195A1

In an embodiment, workflow for timeseries forecasting may be performed based on automated machine learning. Sensor data for measurement parameter is received from plurality of sensors installed in built environment and the received sensor data is stored in table of relational database. Cut-off record associated with previous training checkpoint is determined of the forecasting model for the measurement parameter. Records including new records are determined for which respective timestamps occur after the measurement timestamp of cut-off record. Size of the determined records are compared with threshold size and training dataset is prepared. The forecasting model is trained on the training dataset based on the comparison.

AUTOMATED MACHINE LEARNING FAULT MODELING WITH GROUPING

Absstract of: US20260086257A1

Methods, computing systems, and computer-readable media for a machine learning method of modeling fault-related properties of a geological region are presented. The techniques include: obtaining seismic geological data for a geological region; obtaining from a user identifications of a plurality of faults in the geological region; automatically generating values for descriptors of respective faults of the plurality of faults; automatically partitioning faults of the plurality of faults into a plurality of groups according to the values for the descriptors; obtaining a mapping of respective groups of the plurality of groups to modeling parameter values; applying the mapping to a fault in the geological region outside of the plurality of faults to obtain a modeling parameter value for the fault outside of the plurality of faults; and modeling a fault-related property of the geological region based on the modeling parameter value for the fault outside of the plurality of faults.

IMAGE CLASSIFICATION USING MACHINE LEARNING INCLUDING WEAKLY-LABELED DATA

Absstract of: US20260087618A1

A system may receive a plurality of digital histology images, wherein each of the plurality of digital histology images is labeled with a respective image-level classification. A system may extract a plurality of tiles from each of the plurality of digital histology images. A system may create a first dataset comprising the plurality of tiles and respective image-level classifications. A system may train a first machine learning model using the first dataset. A system may create a second dataset by sampling the first dataset based on respective classifications and respective uncertainty measures for each of the plurality of tiles output by the trained first machine learning model. A system may train a second machine learning model using the second dataset, wherein the trained second machine learning model is configured to classify one or more tiles of a digital histology image.

METHOD FOR SINGLE CELL ANTIMICROBIAL SUSCEPTIBILITY TESTING IN A SUB-DOUBLING TIME

Absstract of: US20260085340A1

Methods and systems for antibacterial susceptibility testing of a bacterium are provided. The method includes exposing a bacterium to an antimicrobial agent. A series of images of the bacterium is captured over time after exposure The series of images are captured during an imaging period. For each image of the series of images, the method includes extracting a value of each feature in a set of morphological features of the bacterium. The set of morphological features includes one or more of area, aspect ratio, length, circularity, perimeter, angularity, curvature, ferret, pole, roundness, sinuosity, width, trajectory, morphology, orientation, solidity, and z-score. A rate of change is calculated for each feature of the set of morphological features during the imaging period. An inhibition status of the bacterium is determined using a machine-learning classifier applied to input data.

MODEL-BASED TASK PROCESSING

Absstract of: US20260087382A1

Embodiments of the disclosure provide a solution for model-based task processing. A method includes: obtaining a base parameter set of a pre-trained base machine learning model, and a first parameter set and a second parameter set of a trained low-rank machine learning model for a first task; applying a Hadamard operator on the base parameter set and the first parameter set, to obtain an intermediate parameter set; aggregating the second parameter set and the intermediate parameter set, to obtain an update parameter set; fine-tuning the base parameter set with the update parameter metric, to obtain a fine-tuned parameter set for a target machine learning model corresponding to the first task; and applying the target machine learning model to perform a model inference for the first task with the fine-tuned parameter set.

BUILDING A DATASET HAVING REQUISITE NUMBER OF FRAUD SAMPLES TO TRAIN A MULTI-CLASS MACHINE LEARNING MODEL FOR FRAUD DETECTION

Absstract of: US20260087380A1

According to an aspect, a system receives a historical data and identifies a set of transactions tagged as fraud (“fraud transactions”) in the received data. If a count of fraud transactions is below a threshold, the system forms a training data and a test data from the historical data, with the test data including all the fraud transactions. The system generates, based on the training data, a one-class anomaly detection model that is able to flag all the fraud transactions when the test data is provided as input to the model. The system applies the model to an inference data to identify whether each transaction therein is an anomaly or not. Upon receiving an input data indicating whether each anomaly is a fraud transaction or not, the system updates the historical data by adding the transactions and tagging the fraud transactions. The updated historical data is used for training a multi-class ML model after the count of fraud transactions is greater than or equal to the threshold.

METHOD IMPLEMENTED IN WASTEWATER MONITORING SYSTEM FOR EVALUATING BIOAVAILABILITY OF ORGANIC NITROGEN IN WASTEWATER

Absstract of: US20260088136A1

A method implemented in a wastewater monitoring system for evaluating bioavailability of organic nitrogen in wastewater, including: obtaining, by using a Fourier Transform Ion Cyclotron Resonance Mass Spectrometer, molecular composition information of organic nitrogen in a wastewater sample collected from a wastewater treatment plant; obtaining bioavailability data corresponding to the wastewater sample, where the bioavailability data is measured through algal bio-culture; training, by a processor of the wastewater monitoring system, a random forest model using the molecular composition information and the bioavailability data; receiving, from the spectrometer, molecular composition information of organic nitrogen in wastewater from a target wastewater treatment plant; and executing, by the processor, the trained machine learning model on the received molecular composition information to generate a predicted bioavailability value; and transmitting, by the wastewater monitoring system, the predicted bioavailability value to a process control unit of the wastewater treatment plant for real-time monitoring or process adjustment.

MACHINE LEARNING TECHNIQUES FOR GENERATING CONTROLLER LOGIC

Absstract of: US20260086524A1

Embodiments of the present disclosure relate to generating controller logic. Indication of a controller logic generation request associated with an asset identifier may be received. A prompt template set associated with a controller logic generation workflow may be identified based on the asset identifier. The prompt template of the prompt template set may comprise one or more instruction sets. The prompt template set may be input into a large language model comprising one or more transformer neural networks and configured to generate a controller logic configuration file for the asset identifier based on the prompt template set and intent classification associated with each prompt template. The controller logic configuration file may be received from the large language model. Performance of one or more prediction-based actions may be initiated based on the controller logic configuration file.

SYSTEMS AND METHODS FOR COLLISION DETECTION

Absstract of: US20260087858A1

A method for detecting vehicle collisions using multi-stage data analysis is described. Telematics data from a vehicle-installed computing device is received and processed through a heuristic filter to identify potential collisions. A feature vector is generated from the filtered data and input into a trained predictive model, which classifies the vector as representing a collision or not. The method then retrieves associated dashcam footage and uses it, along with the predictive model's output, to confirm the occurrence of a collision. Upon confirmation, a notification is transmitted to a remote computing device. This approach combines telematics data analysis, machine learning prediction, and video verification to achieve accurate collision detection and notification.

DATA PRIVACY PROTECTION AND REMOVAL FOR ARTIFICIAL INTELLIGENCE MODEL TRAINING AND DEPLOYMENT

Absstract of: US20260087104A1

There are provided systems and methods for data privacy protection and removal for artificial intelligence model training and deployment. An online transaction processor or other service provider may provide computing services and platforms to entities, which may include use of machine learning (ML) models including large language models (LLMs). To comply with data privacy protections and copyright enforcement, a system may provide unlearning of content from ML models. The system may receive a request to unlearn a content and, after verifying the request is valid, identify the content used for during training of or inferencing by an ML model. The system may then map the content to concepts and correlate those concepts with ML model outputs using projections in a vector space. Based on the mapped concepts and outputs, neuron activation of the ML model may be analyzed to identify a negation vector and perform selective parameter dampening.

METHODS FOR PREDICTING WEIGHT GAIN IN PATIENTS

Absstract of: WO2026064656A1

The present disclosure relates to methods and systems for determining an obesity phenotype in an individual with obesity that utilizes predictive machine learning models and gene risk score calculations. Methods are also provided for selecting responders and non-responders to select pharmacotherapies based on an individual with obesity's determined obesity phenotype as well as methods for predicting weight gain in individuals.

UTILIZING MACHINE LEARNING MODELS TO GENERATE PREDICTED REFEREE INTERACTION METRICS FOR GENERATING AND TRANSMITTING DIGITAL NOTIFICATIONS ACROSS COMPUTER NETWORKS TO REFERRER CLIENT DEVICES

Absstract of: US20260089050A1

0000 The present disclosure relates to systems, non-transitory computer-readable media, and methods for utilizing machine learning to generate predicted referee interaction metrics for building a digital notification distribution policy for tiers of referrer client devices and transmitting digital notifications to referrer client devices across computer networks. In particular, in one or more embodiments, the disclosed systems utilize a referee interaction prediction machine learning model that generate predicted referee interaction metrics indicating likelihoods of downstream interactions of referee client devices based on features of referrer client devices. The disclosed systems generate referrer client device tiers for referrer client devices based on the predicted referee interaction metrics and then utilizes an optimization model to generate a digital notification distribution policy for the tiers of the referrer client devices. Further, the disclosed systems transmit digital notifications to referrer client devices in accordance with the digital notification policy and the referrer client device tiers.

INTER-TECHNOLOGY AND INTER-FREQUENCY MACHINE LEARNING-PROPAGATION SYSTEMS AND METHODS

Absstract of: WO2026064459A1

The devices, systems, and methods described herein are directed to using data associated with a wireless network operational area to train a machine learning model, where the data includes one or more signal characteristic measurement values. The trained machine learning model is used to build a signal propagation model for the wireless network operational area. In some examples, the machine learning model is trained based on signals transmitted in accordance with a first radio access technology (RAT), and the signal propagation model is built for signals transmitted in accordance with a second RAT using the trained machine learning model. In further examples, the machine learning model is trained based on signals transmitted within a first set of frequency bands, and the signal propagation model is built for signals to be transmitted over a second set of frequency bands using the trained machine learning model.

Decision Tree Algorithms in Machine Learning To Learn and To Predict Innovations

Absstract of: US20260087375A1

0000 An expert system for innovation discovery in the field of artificial intelligence (AI) applies decision tree algorithms structured around a rules-based reasoning methodology. The system trains on innovation datasets comprising both data and non-data types, including target variables representing key attributes of innovations and proximal variables that approximate them. Through a machine learning architecture, decision nodes are configured to evaluate these variables and generate predictive models. The architecture enables continual learning via reinforcement mechanisms and communication ports that facilitate data flow from external tools, cloud storage, and software. Differentiated nodes are assigned weights, roles, and activation logic to refine decision-making and improve model accuracy. The expert system integrates human-defined heuristic rules with AI capabilities to support early-stage ideation, concept development, and innovation pattern recognition. This system can operate as an autonomous AI agent, a core reasoning engine, or as part of a digital business model in platform ecosystems to enhance innovation discovery, reduce hallucinations, and support transparent, verifiable AI outcomes.

DEEP NEURAL NETWORKS (DNN) INFERENCE USING PRACTICAL EARLY EXIT NETWORKS

Nº publicación: US20260086912A1 26/03/2026

Applicant:

MICROSOFT TECH LICENSING LLC [US]
Microsoft Technology Licensing, LLC

Absstract of: US20260086912A1

The present disclosure relates to methods and systems for providing inferences using machine learning systems. The methods and systems receive a load forecast for processing requests by a machine learning model and split the machine learning model into a plurality machine learning model portions based on the load forecast. The methods and systems determine a batch size for the requests for the machine learning model portions. The methods and systems use one or more available resources to execute the plurality of machine learning model portions to process the requests and generate inferences for the requests.