Alerta

COMPUTER-BASED SYSTEMS CONFIGURED FOR ENTITY RESOLUTION AND INDEXING OF ENTITY ACTIVITY

Resumen de: US20260278418A1

In order to facilitate the entity resolution and entity activity tracking and indexing, systems and methods include receiving first source records from a first database and second source records from a record database. A candidate set of second source records is determined by a heuristic search in the set of second source records. A candidate pair feature vector associated with each candidate pair of first and second source records is generated. An entity matching machine learning model predicts matching first source records for each candidate second source record based on the respective candidate pair feature vector. An aggregate quantity associated with the matching first source records is aggregated from a quantity associated with each first source record, and a quantity index for each candidate second source record is determined based the aggregate quantities. Each quantity index is displayed to a user.

MACHINE LEARNING-ENABLED FEEDBACK

Resumen de: US20260278650A1

Methods disclosed herein include receiving a record of an interaction between an agent and a human participant in the interaction, analyzing the record to generate at least one characteristic of the interaction attributable to the human participant, wherein the characteristic comprises: a topic, a non-word symbol, metadata, a period of silence, a language characteristic, or an acoustic aspect, determining at least one category or score for at least a portion of the interaction based on the characteristic, generating a summary of the interaction using a machine learning model, and generating a customized communication based on the category or score, wherein the customized communication includes the summary and at least one prompt for the human participant.

Machine Learning Techniques For Generating Skill Hierarchies

Resumen de: US20260278493A1

0000 Techniques for using machine learning to generate skill hierarchies are disclosed. A system generates a graph that models a set of skills, including first and second nodes that represent first and second subsets of the skills. The system uses a model to determine that the first and second nodes are candidates to be connected by an edge in the graph, then uses a large language model to determine that the nodes are sufficiently similar to be connected by the edge. The system performs a minimum cut of a connected subgraph that includes the first and second nodes, to obtain first and second connected sub-subgraphs. The system uses generative artificial intelligence to determine (a) first and second node-level representatives for the first and second nodes and (b) a sub-subgraph-level representative skill. The system generates a skill hierarchy that includes the node-level and sub-subgraph-level representative skills.

SYSTEM, METHOD, AND COMPUTER PROGRAM PRODUCT FOR OPTIMIZING MACHINE LEARNING MODELS USING GRAPH-BASED EMBEDDINGS

Resumen de: WO2026192923A1

Systems, methods, and computer program products are provided for optimizing machine learning models using graph-based embeddings. An example system includes a processor to generate a graph structure, generate a plurality of graph embeddings based on neighboring nodes of a node in the graph structure, generate transformed embeddings by inputting a graph embedding to a neural network to generate an output, the output including one or more transformed embeddings that correspond to the graph embedding, combine the transformed embeddings with trainable intermediate embeddings of a machine learning model to provide an intermediate input for the machine learning model, input the intermediate input to the machine learning model, and perform a training procedure on the plurality of graph embeddings, the neural network, the trainable intermediate embeddings, and the large foundational machine learning model based on an output of the machine learning model that was generated based on the intermediate input.

Systems and Methods for Dynamic Control of Machine Learning Parameters

Resumen de: US20260278469A1

Systems and methods for dynamic control of machine learning parameters, allowing for modification of weights for input values and vectors during processing. Implementations of these systems provide greater flexibility without requiring extensive retraining of a model, and may improve accuracy of classifications for diagnostic systems or other analyzers.

MACHINE LEARNING SYSTEM, SERVER DEVICE, CLIENT DEVICE, MACHINE LEARNING METHOD, AND COMPUTER READABLE MEDIUM

Resumen de: US20260278420A1

A machine learning system (500) includes a client device (200) that has its own unique learning data and a server device (100) that has training data. Each of the client device (200) and the server device (100) has an AI model. Further, the machine learning system (500) proceeds with learning while keeping the learning data of the client device (200) and the training data of the server device (100) private to each other. The server device (100) includes a fraudulent client detection unit (2530) to estimate a fraudulent client device that performs an attack to disrupts the learning of the AI model in the server device (100).

APPARATUS, METHOD, AND SYSTEM FOR MACHINE LEARNING JOB CHECKPOINTING IN TRUSTLESS DISTRIBUTED TRAINING PLATFORMS

Resumen de: US20260278400A1

An approach is provided for Machine Learning (ML) job checkpointing in trustless distributed training platforms. The approach, for example, involves receiving an abort message from a resource owner system. The abort message relates to a ML training job that is running. The approach also involves receiving a checkpoint claim from the resource owner system based on the cancelling of the ML training job. The approach further involves receiving a proof of the checkpoint claim comprising at least one of: a proof of execution of the ML training job generated within a Trusted Execution Environment (TEE); or a creation of a future restore failure transaction on a Distributed Ledger Technology (DLT) network configured to execute under a dispute protocol concerning a validity of the checkpoint claim. The approach further involves causing, at least in part, a storage of training artifacts of the ML training job based on the proof.

SYSTEMS AND METHODS FOR CLUSTERING AND SEPARATING GRAPH NETWORKS

Resumen de: US20260278467A1

Systems and methods for decomposing an optimization problem to reduce solve time. Input data associated with the optimization problem is read and the problem is represented as a graph comprising nodes and edges. A machine learning technique clusters the nodes into a plurality of clusters, and an association between constraints and edges spanning the clusters is obtained. For each constraint, a distribution of predicted load among the clusters is determined based on demand information generated by a machine learning model trained on the input data. At least one constraint is then split across the plurality of clusters in accordance with the predicted load distribution.

GENERATING CREDIT CAPACITY ESTIMATES

Resumen de: US20260278675A1

One or more computer processors are used to generate an estimate of a credit capacity of a client. Financial data relating to the client is received. The financial data is input to one or more trained machine learning models. The one or more trained machine learning models are used to generate, based on the financial data, the estimate of the credit capacity.

EXPLAINING RELATIONAL FEATURES UTILIZING SURROGATE ARTIFICIAL INTELLIGENCE MODELS

Resumen de: US20260281140A1

0000 Approaches to explaining relational features utilizing surrogate artificial intelligence (AI) models are described. An input to a machine learning (ML) model is received. The ML model returns a verdict corresponding to a network security evaluation-based graph classification. A surrogate AI model is trained to minimize a loss function while causing a masked output to come as close as possible numerically to the graph classification generated without a mask involved in a mathematical equation of the ML model. The AI model utilizes multiple different kinds of masks. Rating reports are generated based on learned masks from the surrogate AI model to indicate nodes or edges in the ML model most heavily responsible for the verdict returned from the ML model.

SYNTHETIC DATA GENERATION FOR APPLICATION PROGRAMMING INTERFACE REQUEST PAYLOADS AND MOCK DEPENDENCY RESPONSES

Resumen de: US20260277709A1

Systems and methods for generating application programming interface (API) request payloads and mock dependency responses. In some aspects, the system receives an API request schema for an API and a target output field and corresponding value for the API. The system generates, using a machine learning optimization algorithm, a plurality of candidate solutions. The system generates an API request to the API with each candidate solution to obtain one or more corresponding output values for the target output field. The system determines a fitness score of each candidate solution. The system adjusts, using the machine learning optimization algorithm, based on corresponding fitness scores, the plurality of candidate solutions. The system iteratively repeats generating the API request, determining the fitness score of each candidate solution, and adjusting the plurality of candidate solutions until the machine learning optimization algorithm converges on a final candidate solution.

SYSTEM

Resumen de: US20260278341A1

A system includes a processor that is configured to collect data from a plurality of information sources on the Internet, perform natural language processing on the collected data to analyze the collected data and extract important keywords and phrases from the collected data, evaluate a reliability of information included in the collected data by using a deep learning algorithm, provide evaluated information to a user, evaluate the reliability of the information by taking into account at least a source of the information, consistency of content of the information, a past reliability evaluation of the information or of the source, and an update frequency of the information, provide, to the user, information that is determined to have high reliability without modification, and display a warning for information that is determined to have low reliability and provide a detailed explanation regarding the reliability of the information.

"AN IMPROVED SYSTEM FOR EXTUBATION TYPE PREDICTION"

Resumen de: WO2026190804A1

The present invention relates to an improved system for extubation type predictions in patients undergoing mechanical ventilation for various respiratory or neurological health conditions The system comprises a 5 data acquisition and pre-processing unit (100), a model training and evaluation unit (200), a model deployment and prediction unit (300), and a clinical decision support unit (400). Furthermore, the system continuously collects, processes, and analyzes patient data to minimize medical risks associated with premature or delayed extubation, aiming 10 for improved patient outcomes in critical care settings. Moreover, the system uses advanced machine learning techniques to improve the accuracy and dependability of extubation predictions in real-time, helping medical professionals make well-informed decisions, which leads to better patient safety, fewer complications, and efficient resource 15 utilization.

Systems and Methods for Inferring Biological, Environmental, Ecological, and Therapeutic States from Optical Interaction Dynamics and Multivariable Longitudinal Monitoring

Resumen de: US20260279003A1

Systems and methods are provided for inferring biological, environmental, ecological, therapeutic, industrial, agricultural, or physiologic states from optical interaction signatures and temporal optical interaction dynamics obtained through multivariable longitudinal monitoring. One or more imaging devices or sensors acquire optical interaction data associated with biological matrices, environmental samples, atmospheric systems, industrial processes, or other monitored environments. One or more processors extract quantitative optical interaction features—including Global Edge Coherence (GEC), Localized Edge Dynamics (LEDyn), speckle evolution, aggregation kinetics, and related metrics—and integrate these features with environmental variables, geospatial data, biologic measurements, therapeutic history, and longitudinal multimodal datasets using machine-learning and probabilistic inference engines. In certain embodiments, the system receives and processes structured Digital Twin Manifests (DTMs). The resulting probabilistic inferences support closed-loop adaptive monitoring, real-time therapeutic-response assessment, distributed smartphone-based environmental intelligence, environmental digital biomarkers, and proactive recommendations across laboratory, clinical, ecological, manufacturing, and extraterrestrial applications.

MODEL ECOSYSTEM

Resumen de: US20260278425A1

A system of machine learning (“ML”) models for making actionable predictions regarding low-incidence events, including a generative ML model that produces synthetic minority-class records to form an augmented training data set, a predictive ML model that has been trained on the augmented training data set, a certainty ML model that produces a certainty estimate, and an explanatory model that produces an explanation. A method for producing actionable predictions of a low-incidence event by applying ML models to imbalanced class data by producing a prediction by a predictive ML model that has been trained on a data set comprising synthetic minority-class data records produced by a generative ML model, and producing a certainty estimate and an explanation. At least one of the certainty estimate or explanation determines an effective or appropriate response to the prediction. The low-incidence event may comprise risk of opioid use disorder.

MODEL MANAGEMENT METHOD AND APPARATUS

Resumen de: US20260278419A1

A model management method and an apparatus, and relates to the field of communication technologies, to reduce a data security risk of an AI model in the wireless field. The method includes: obtaining a first communication data set used for updating a first machine learning model, and sending a second communication data set to a model training function entity. The first processing policy includes a first loss threshold, and the first machine learning model is used for managing a wireless communication service. The second communication data set is obtained by performing a processing operation on suspicious data in the first communication data set according to the first processing policy, and the suspicious data is wireless communication data whose loss value is greater than the first loss threshold.

EARLY WARNING AND COLLISION AVOIDANCE

Resumen de: US20260279201A1

Among other things, equipment is located at an intersection of a transportation network. The equipment includes an input to receive data from a sensor oriented to monitor ground transportation entities at or near the intersection. A wireless communication device sends to a device of one of the ground transportation entities, a warning about a dangerous situation at or near the intersection, there is a processor and a storage for instructions executable by the processor to perform actions including the following. A machine learning model is stored that can predict behavior of ground transportation entities at or near the intersection at a current time. The machine learning model is based on training data about previous motion and related behavior of ground transportation entities at or near the intersection. Current motion data received from the sensor about ground transportation entities at or near the intersection is applied to the machine learning model to predict imminent behaviors of the ground transportation entities. An imminent dangerous situation for one or more of the ground transportation entities at or near the intersection is inferred from the predicted imminent behaviors. The wireless communication device sends the warning about the dangerous situation to the device of one of the ground transportation entities.

SYSTEM, RFID CHIP, SERVER AND METHOD FOR CAPTURING VEHICLE DATA

Resumen de: US20260279526A1

0000 A system includes a plurality of tracking devices, such as RFID tags, affixed to items, such as vehicles, a data collection engine, client devices and backend devices. The backend devices include trained machine learning models, business logic, and attributes of a plurality of events. A plurality of data collection engines and systems send attributes of new events to the backend devices. The backend devices can track the items and predict particular outcomes of new events based upon the attributes of the new events utilizing the trained machine learning models.

MODELED CONVERSIONS FOR A CONNECTION NETWORK SYSTEM

Resumen de: US20260278339A1

Modeled conversions for connection networking are described. A method comprises generating a first input vector comprising a set of sequence features for a group of entities using a transformer layer of a machine learning model, generating a second input vector comprising a set of identity profile features for the group of entities using a generative artificial intelligence layer, receiving a third input vector comprising a set of conversion features, concatenating the first input vector, the second input vector, and the third input vector into a concatenated vector, normalizing the concatenated vector to form a normalized vector, projecting the normalized vector to form a projected vector, and generating a first metric based on the projected vector using a multi-gate mixture-of-experts (MMOE) layer of the ML model, the first metric comprising a value indicating whether there is a deterministic entity level conversion event for a group level conversion event.

RESOURCE PREDICTION FOR QUANTUM ADVANTAGE ON THE USAGE OF GATED QUANTUM COMPUTERS

Resumen de: US20260278444A1

0000 One example method includes selecting a set of classical computing algorithms, solving, using the classical computing algorithms, respective problem instances of a problem to obtain a set of executions, where respective ML (machine learning) models were trained using the executions of the problem instances to predict a respective amount of classical computing resources needed to execute a respective one of the problem instances by each of the classical computing algorithms, in a loop, using an ensemble ML model to predict the classical resources needed to solve the problem, and when the predicted classical resources exceed defined classical resource limits, ending the loop, constructing a quantum circuit that comprises, as among a respective set of gates and qubits of quantum algorithms in a group of quantum algorithms, a smallest set of gates and qubits, and determining minimum requirements of quantum hardware needed to run the quantum circuit with a specified fidelity.

ADAPTIVE ALIGNMENT TRANSFER FOR MACHINE LEARNING MODELS

Resumen de: US20260277951A1

0000 At least one processor may generate at least one modification for at least one weight of an unaligned target large language model (LLM). The generating may include obtaining at least one delta defining a difference between at least one layer of an unaligned source LLM and the at least one layer of an aligned source LLM and performing at least one mutation operation on the at least one delta. The at least one processor may inject the at least one modification into at least one target layer of the unaligned target LLM, thereby generating an aligned target LLM.

MATERIAL STRUCTURE AND SYNTHESIS PATHWAY GENERATION USING GENERATIVE MODELS

Resumen de: WO2026193491A1

A method for designing a synthesis pathway for a target material is provided herein. The method includes receiving input data related to a target material. The method further includes analyzing the input data using a machine learning model to generate one or more candidate synthesis pathways. The method further includes selecting a synthesis pathway from the candidate synthesis pathways. The method further includes outputting the selected synthesis pathway design.

SYSTEM AND METHODS FOR ONE-SHOT FEDERATED UNLEARNING

Resumen de: WO2026188482A1

A method for training a machine learning model in a federated learning system is disclosed. The federated learning system includes a plurality of client devices and a central server that initializes a global model and aggregates locally trained updates from the client devices to update the global model. The global model and a forgetting threshold are provided to a forget client. The forget client calculates a Fisher information matrix diagonal, and generates a network mask using the Fisher information matrix diagonal and the forgetting threshold. The network mask is an indicator of specialized parameters of the global model for the local training data of the forget client. The forget client transmits the network mask to the central server or all other clients of the federated learning system. The central server or the other clients randomly initialize or set the specialized parameters of the global model.

SYSTEMS AND METHODS FOR AUTOMATED EXIT VERIFICATION

Resumen de: US20260278577A1

0000 Systems and methods provided herein may facilitate store exit verification. Methods may include capturing, in real-time and as a customer approaches an exit, a set of images of one or more items. Methods may include constructing, via a computer vision engine, a preliminary item catalog based on the one or more items detected near the exit. Methods may include constructing, via a machine learning engine and based on the preliminary item catalog, a refined item catalog by cross-referencing attributes of the one or more items against at least one of synthetic data, historical purchase patterns, contextual product groupings, and known inventory information. Methods may include detecting, via a verification engine and in real-time, discrepancies between the refined item catalog and a corresponding transaction log obtained from a transaction database, wherein the customer is allowed to pass through the exit based on the discrepancies.

TECHNOLOGIES FOR MACHINE LEARNING CLOUD BASED CLINICAL OUTCOME PREDICTION

Nº publicación: US20260279577A1 17/09/2026

Solicitante:

CARE PREDICTOR LLC [US]
Care Predictor LLC

Resumen de: US20260279577A1

0000 Technologies for clinical outcome predictive analysis include a computing device that extracts provider features from an assessment index that are indicative of relational characteristics of a care provider. The computing device determines a clinical outcome performance metric associated with the care provider. The computing device links the provider features and the performance metric in training data and trains a prediction model to predict an outcome score based on provider feature. A computing device may predict a predicted outcome score based on extracted provider features using the trained prediction model and normalize the predicted outcome score to a predicted outcome index. The computing device may monitor ongoing therapeutic alliance status data and predict a risk of alliance failure using a trained prediction model. Other embodiments are described and claimed.