Alerta

FEDERATED MACHINE LEARNING-BASED ANOMALY DETECTION ACROSS MULTIPLE DATA CENTERS

Resumen de: US20260289386A1

An apparatus comprises at least one processing device configured to generate graph structure representations of a first set of logs associated with information technology assets in a first data center, to train a first machine learning-based anomaly detection model utilizing the graph structure representations, to train a second machine learning-based anomaly detection model by distilling knowledge of the first machine learning-based anomaly detection model, and to update the second machine learning-based anomaly detection model based on an aggregation of parameters of the second machine learning-based anomaly detection model and additional parameters of additional machine learning-based anomaly detection models trained utilizing logs associated with information technology assets of additional data centers. The at least one processing device is also configured to perform anomaly detection on a second set of logs associated with the information technology assets in the first data center utilizing the updated second machine learning-based anomaly detection model.

STACKING OF HETEROGENEOUS MACHINE LEARNING MODELS FOR DETECTION OF FRAUDULENT TRANSACTIONS

Resumen de: US20260289568A1

A system is adapted to automatically detect fraudulent transactions. The system includes a fraud management computer system programmed for: receiving a plurality of transactions, some legitimate and some fraudulent; with the transactions and a K-fold training module, training a heterogeneous group of machine learning models to classify the transactions as legitimate or fraudulent, where the K-fold training module generates, for each machine learning model, a mean and a standard deviation; with the means and standard deviations, determining a weight for each machine learning model; and, with the weights for each machine learning model, defining a meta machine learning model to classify the transactions as legitimate or fraudulent; receiving a real-time transaction and, with the trained machine learning models, generating determinations of whether it is legitimate or fraudulent; with the meta machine learning model and the determinations, classifying the real-time transaction as legitimate or fraudulent; and, if fraudulent, generating an alert.

EFFICIENT SPECULATIVE DECODING USING DRAFTING TREES FOR MACHINE LEARNING MODELS

Resumen de: WO2026193870A1

Certain aspects of the present disclosure provide techniques and apparatus for machine learning. In an example method, a drafting tree comprising a first plurality of draft tokens is generated using a generative machine learning model. A plurality of confidence scores comprising a respective confidence score for each respective token of the first plurality of draft tokens is determined. A plurality of acceptance predictions comprising a respective acceptance prediction for each respective token of the first plurality of draft tokens is generated based on the plurality of confidence scores and a structure of the drafting tree, and a subset of draft tokens is selected from the first plurality of draft tokens based on the plurality of acceptance predictions. An output token is generated based on processing the subset of draft tokens using the generative machine learning model.

AUTOMATED VALUATION MODEL (AVM) AND TECHNIQUES FOR REDUCING LEAKAGE IN MACHINE LEARNING DATASETS

Resumen de: US20260289411A1

0000 System, apparatus and method may control receiving time-indexed data comprising values of a plurality of features; generating derived features based on temporal changes of at least one of the plurality of features, the derived features including at least a first derivative of a first value; temporally aligning the plurality of features and the derived features to a reference date such that only data available prior to the reference date is used for training a machine learning model; training a machine learning model on the temporally aligned features and the derived features to predict a forward-looking value; and outputting the forward-looking value.

ARTIFICIAL INTELLIGENCE BASED RULE GENERATION FOR DATABASE CHANGE DEPLOYMENT

Resumen de: US20260288683A1

0000 Embodiments provide systems, methods, and computer program products that utilize artificial intelligence/machine learning to process database change data and correlated performance data to predict the impact of database changes and generate rules with respect to database changes to prevent undesired behavior or promote increased performance.

MACHINE LEARNING TRAINING DURATION CONTROL

Resumen de: US20260289413A1

Some embodiments select a machine learning model training duration based at least in part on a fractal dimension calculated for a training data dataset. Model training durations are based on one or more characteristics of the data, such as a fractal dimension, a data distribution, or a spike count. Default long training durations are sometimes replaced by shorter durations without any loss of model accuracy. For instance, the time-to-detect for a model-based intrusion detection system is shortened by days in some circumstances. Model training is performed per a profile which specifies particular resources or particular entities, or both. Realistic test data is generated on demand. Test data generation allows the trained model to be exercised for demonstrations, or for scheduled confirmations of effective monitoring by a model-based security tool, without thereby altering the model's training.

ENSEMBLE MACHINE LEARNING SYSTEM FOR DYNAMIC RESOURCE ALLOCATION USING CROSS-SOURCE DATA AGGREGATION

Resumen de: US20260289661A1

A computer-implemented method for evaluating resource allocation requests using an ensemble machine learning model is disclosed. A resource request is received, comprising multiple data inputs from different data sources, where each data source provides a specific type of user feature. A first set of machine learning models, each trained on a single data source, generates corresponding first risk scores. A second decision tree-based machine learning model processes the data inputs, leveraging feature dependencies and learned missing feature patterns to generate a second risk score, enabling robust predictions even when certain user features are incomplete or missing. The first and second risk scores are ensembled using an ensemble technique to produce an overall risk score. The ensembled risk score is then transmitted to a resource hosting device, which approves, denies, or adjusts the requested resource based on the assessed risk.

METHODS, SYSTEMS AND COMPUTER READABLE MEDIA FOR TRAINING MACHINE LEARNING MODELS

Resumen de: WO2026197916A1

A method for training a machine learning model, comprising determining a plurality of input data objects from a first unlabelled set of data, and providing the plurality of input data objects to at least one reference model configured to generate a first output for each of the plurality of input data objects. A labelled set of training data is created using the generated first outputs, and a target model is trained using the labelled set of training data. The trained target model is applied to a second unlabelled set of data. A plurality of training data objects is determined from the second unlabelled set of data. The plurality of training data objects is provided as an input to the at least one reference model to generate a second output for each of the plurality of training data objects, and the labelled set of training data is updated.

AUTONOMOUS GENERATIVE INTELLIGENCE ENVIRONMENT METHOD ACROSS OPERATING SYSTEM, DEVICES AND CONNECTED NETWORKS.

Resumen de: WO2026196276A2

A computer-implemented system and method for operating a local cross-application autonomous task routing framework tangibly embodied as a low-level software operational layer designated as 49AGI app via playstore and ios app store. The system architecture intercepts multi-modal input data directly from a plurality of distinct local applications, bypasses visual screen-pixel processing, and manages active task states natively via local processor networks. The 49AGI architecture executes background operations across applications lacking native machine learning models, providing local multi-modal reminder generation, notification scheduling, and native document compilation.

VALIDATION OF RESPONSES GENERATED BY LANGUAGE MODELS

Resumen de: US20260289392A1

Validation of responses generated by a language model is provided and includes identifying, via machine learning, first entity action pairs from a first response of a generative language machine learning model. Each of the first entity action pairs includes a respective entity and a respective action associated with the respective entity. A first validation score is determined based on whether the first entity action pairs are present within a first knowledge base and a second set of validation scores is determined based on whether the first entity action pairs are present within a second knowledge base. A first final validation score is generated based on the first validation score and the second set of validation scores. A first recommendation from the first response is presented based on the first final validation score and operations are controlled based on the presented first recommendation.

MACHINE LEARNING ENABLED CONFORMATION-AWARE PROTEIN STRUCTURE PREDICTION

Resumen de: US20260290493A1

0000 A method for conformation aware protein structure prediction may include receiving an amino acid residue sequence of an input protein molecule, such as at least a portion of an antibody, a nanobody, or a T-cell receptor (TCR). A representation of the amino acid residue sequence of the input protein molecule may be determined to include a conformation token specifying a conformation state of the input protein molecule. The conformation state of the input protein molecule may be an unbound state or a bound state. A protein structure computation model may be applied to determine, based on the representation of the amino acid residue sequence of the input protein molecule, a conformation of the input protein molecule in the conformation state specified by the conformation token. Related systems and computer program products are also provided.

GENERATING TIMELINES USING MACHINE LEARNING MODELS

Resumen de: WO2026198917A1

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a timeline. One of the methods includes obtaining a plurality of documents; generating, for each document, a corresponding annotated document; generating, for each corresponding annotated document, an embedded annotated document; generating, for each document, a corresponding timeline for the document; and generating a combined timeline over the plurality of documents by merging each of the corresponding timelines based on at least the embedded annotated document for one or more of the documents.

COMPUTER-IMPLEMENTED SYSTEM AND METHOD FOR ADAPTIVE BEHAVIORAL MODELING AND CONTENT OPTIMIZATION USING ARTIFICIAL INTELLIGENCE TECHNIQUE

Resumen de: US20260289358A1

A method and system for simulating responses from virtual personas, the method comprising: using a computer processor, generating a plurality of target audience descriptions; converting the plurality of target audience descriptions into numerical embeddings; converting a pre-existing personality survey into numerical embeddings; determining, with a machine learning model, compatibility of the plurality of target audience descriptions with respondents of the pre-existing personality survey; forming an array of likelihood scores based on the compatibility, wherein columns of the array correspond to the target audience descriptions and rows correspond to the pre-existing personality survey; selecting a subset of the rows of the array with likelihood scores above a user-selected threshold; generating, from the subset, a set of virtual personas combining the plurality of target audience descriptions and the pre-existing personality survey respondents; and simulating responses using the set of virtual personas.

CROSS SPLINE NETWORK

Resumen de: US20260289281A1

A computer system for machine learning-based prediction is disclosed. The system applies a neural network-based spline transformation to input variables to generate transformed features. The transformed features are processed using a cross network comprising one or more cross layers, each configured to model feature interactions by combining prior layer outputs with learned weight parameters and applying a “cross” (multiplication) transformation with the input features. An output prediction is generated based on the transformed features processed by the cross network. The disclosed approach enables improved modeling of feature interactions while leveraging neural network-based transformations, facilitating enhanced interpretability and computational efficiency in machine learning applications.

Systems and Methods for Adaptive Avatar Dialog Generation and Behavioral Modeling

Resumen de: US20260284519A1

0000 A deep-learning system enables a coach to guide interactions between a user and an avatar while adaptively refining behavioral characteristics associated with the avatar. During a session, a coach provides prompts, feedback, and scoring through a dedicated interface, and the system generates avatar-matched dialog based on predefined behavioral traits, objectives, and personality attributes. The system compares automated scoring with coach-provided evaluations to identify substantive differences and dynamically update adaptive model parameters used in future dialog generation. The invention further supports real-time response suggestions, behavioral reinforcement, avatar movement controls, voice transformation, and personality-based language modification. Training data generated during sessions may be reviewed and incorporated into ongoing deep-learning processes, enabling continuous improvement of avatar behavior modeling for coaching, simulation, education, therapeutic, or interactive communication environments.

ADAPTIVE TRANSMISSION AND TRANSMISSION PATH SELECTION BASED ON PREDICTED CHANNEL STATE

Resumen de: US20260292652A1

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a transmitter node may predict a future state associated with a wireless channel at a future time instance using a machine learning model, wherein the future state is predicted based at least in part on one or more of weights associated with the machine learning model, a current state associated with the wireless channel, or one or more previous states associated with the wireless channel. The transmitter node may select one or more parameters for a transmission to occur at the future time instance based at least in part on the future state associated with the wireless channel. The transmitter node may perform the transmission using the one or more parameters. Numerous other aspects are described.

SMART RING SYSTEM FOR MEASURING DRIVER IMPAIRMENT LEVELS AND USING MACHINE LEARNING TECHNIQUES TO PREDICT HIGH RISK DRIVING BEHAVIOR

Resumen de: US20260285324A1

A method for predicting risk exposure can include sensing, by a sensor of a ring worn by a user, a set of data. The sensor can be integrated with a microfluidic device. The method for predicting risk exposure further can include determining, by a trained machine learning (ML) model, a risk score based at least on the set of data. The method for predicting risk exposure also can include determining, by a trained ML model, a remediating action to reduce the risk score. The method for predicting risk exposure also can include, in response to determining the risk score, generating a notification to alert the user of the risk score and the remediating action. Other embodiments are disclosed.

HYBRID QUANTUM CALCULATING SYSTEM USING ARTIFICIAL INTELLIGENCE AND OPERATION METHOD THEREOF

Resumen de: US20260289381A1

A hybrid quantum calculating system according to the present disclosure comprises a correction information acquisition device acquiring correction information for correcting an error of at least one of a first qubit and a second qubit based on first environment information and second environment information, and a qubit error correction device correcting values of a first qubit register and a second qubit register based on the correction information. Also, the hybrid quantum calculating system acquires module selection information by applying environment information to a dynamic selection machine learning model, and selects at least one of a first encryption module, a second encryption module, and a third encryption module based on the module selection information.

LANGUAGE-BASED CROSS-MODALITY TRANSFER SYSTEM FOR GENERATING VIRTUAL DATA TO ENHANCE ACTIVITY RECOGNITION

Resumen de: US20260289407A1

0000 An exemplary system and method are disclosed for generating training data for artificial intelligence (AI) or machine learning (ML) models to detect or recognize activities of a subject. In one implementation, an example method includes generating, via a motion synthesis model and as a first trained AI model, a set of motion sequence data objects representing motion sequences corresponding with the plurality of activity descriptions, where the generation of the set of motion sequence data objects includes applying an iterative data-variability evaluation operation.

READER LIFE SPAN PREDICTION MODEL

Resumen de: WO2026195627A1

A system for predicting lifespan of a physical access control (PAC) reader device is described. The system collects data associated with a PAC reader. The system processes the data using one or more machine learning (ML) models to generate a prediction of a remaining lifespan of the PAC reader and one or more configuration changes for extending the remaining lifespan of the PAC reader. The system generates an alert comprising the prediction and the one or more configuration changes.

INTEGRATED TELEMETRY PIPELINE

Resumen de: US20260289286A1

0000 Embodiments described herein are directed to a multi-layer telemetry pipeline utilizing Machine Learning (ML) and/or Artificial Intelligence (AI) to perform reduction-based telemetry collection for anomaly detection and/or prediction tasks. Throughout the pipeline the amount of data is reduced at each level. As telemetry pipelines usually collect many data items into a central collector, through layers of aggregation, ML/AI can be integrated into each layer, i.e., the pipeline nodes/layers themselves to execute layers of the algorithm. In other words, the “low-level”, e.g. feature extraction layers of the pipeline can be executed within a local aggregator. This reduces the data to be sent, making it easier but without losing data to be ingested by the algorithm. The “higher level,” result-calculating parts of the algorithm can be executed within a collector.

OPTIMIZING ARTIFICIAL INTELLIGENCE GENERATED CODE ON THE COMPUTING CONTINUUM

Resumen de: US20260288429A1

Systems and methods for optimizing artificial intelligence generated code on the computing continuum. An instruction code that instructs a machine learning model (MLM) can be modified to obtain a modified instruction code that incorporates a dynamic control flow that limits processing of software code based on a target time period. One or more candidate codes based on the modified instruction code can be generated. One or more service paths within the dynamic control flow of the one or more candidate codes can be executed to obtain an optimized generated code having detailed responses within a threshold for downstream tasks by asynchronously performing sub-tasks from the one or more service paths to optimal computing nodes.

Machine Learning Approach to Multi-Domain Process Automation and User Feedback Integration

Resumen de: US20260289418A1

Embodiments relate to multi-domain process automation with user feedback integration. Some embodiments include a method performed by one or more computing devices. The one or more computing devices generate, using a machine learning (ML) model, predictions for records. The one or more computing devices receive at least one of single user feedback or multiple user feedback for the predictions. The one or more computing devices generate a user validated record pool based on the single user feedback or multiple user feedback. The one or more computing devices update the ML model using the user validated record pool.

SYSTEM AND METHOD FOR CALCULATING USER SATISFACTION INDICES BASED ON MULTIMODAL DATA

Resumen de: WO2026198071A1

A non-invasive, privacy-conscious system and method for calculating User Satisfaction Index (USI) in real-time using multi-sensory data fusion. The system is primarily based on radar sensors, ensuring privacy, while optical and acoustic sensors are optional and can be integrated as needed. It employs machine learning algorithms to analyze physiological (heart rate, respiration, micro-movements), emotional response, and behavioral (gestures, vocal intonation, facial expressions) markers. The User Satisfaction Index prediction and recommendation algorithms estimate users' satisfaction across various industries, including retail, healthcare, corporate environments, smart cities, transportation, and other fields.

METHODS, SYSTEMS, INTERFACES AND COMPUTER READABLE MEDIA FOR TRAINING MACHINE LEARNING MODELS

Nº publicación: WO2026197915A1 24/09/2026

Solicitante:

XERO LTD [NZ]
XERO LIMITED

Resumen de: WO2026197915A1

A method for extracting data from a data object comprising receiving, at an extraction application programming interface (API), an extraction request comprising at least one data object or at least one data object reference and an extraction schema comprising one or more target data elements. The method includes determining for each target data element, whether a target model exists for extracting the target data element. Responsive to determining that a target model exists, the target model is applied to the data object to extract one or more values associated with the target data element from the data object. Responsive to determining that a target model does not exist, a general extractor model is applied to the data object to extract one or more values associated with the target data element from the data object. The method further includes generating and outputting an extraction payload.