Alerta

DYNAMIC FILTER RECOMMENDATIONS

Resumen de: US2025181587A1

A user preference hierarchy is determined from user response to images. Images may be tagged using machine learning models trained to determine values for images. Products are clustered according to product vectors. Images of products within a cluster are clustered according to composition and groups of images are selected from image clusters for soliciting feedback regarding user preference for products of a cluster. Feedback is used to train a user preference model to estimate affinity for a product vector. A user may provide feedback regarding a price point and products are weighted according to a distribution about the price point. The distribution may be asymmetrical according to direction of movement of the price point. Filters may be dynamically defined and presented to a user based on popularity and frequency of occurrence of attribute-value pairs of search results and based on feedback regarding the search results.

SYSTEMS AND METHODS FOR PROGRAMMATIC LABELING OF TRAINING DATA FOR MACHINE LEARNING MODELS VIA CLUSTERING AND LANGUAGE MODEL PROMPTING

Resumen de: AU2023383086A1

Embodiments introduce an approach to semi-automatically generate labels for data based on implementation of a clustering or language model prompting technique and can be used to implement a form of programmatic labeling to accelerate the development of classifiers and other forms of models. The disclosed methodology is particularly helpful in generating labels or annotations for unstructured data. In some embodiments, the disclosed approach may be used with data in the form of text, images, or other form of unstructured data.

MACHINE LEARNING BASED OCCUPANCY FORECASTING

Resumen de: WO2025117106A1

Embodiments determine a final occupancy prediction for a check-in date for a plurality of hotel rooms. Embodiments receive historical reservation data including a plurality of booking curves for the hotel rooms corresponding to a plurality of reservation windows, the historical reservation data including a plurality of features. Based on the historical reservation data, embodiments generate a first occupancy prediction for the check-in date using a first model and generate a second occupancy prediction for the check-in date using a second model. Embodiments determine a best performing model from at least the first model and the second model uses a corresponding occupancy prediction corresponding to the best performing model as the final occupancy prediction for the check-in date.

CONCURRENT OPTIMIZATION OF MACHINE LEARNING MODEL PERFORMANCE

Resumen de: US2025181978A1

Certain aspects of the present disclosure provide techniques for concurrently performing inferences using a machine learning model and optimizing parameters used in executing the machine learning model. An example method generally includes receiving a request to perform inferences on a data set using the machine learning model and performance metric targets for performance of the inferences. At least a first inference is performed on the data set using the machine learning model to meet a latency specified for generation of the first inference from receipt of the request. While performing the at least the first inference, operational parameters resulting in inference performance approaching the performance metric targets are identified based on the machine learning model and operational properties of the computing device. The identified operational parameters are applied to performance of subsequent inferences using the machine learning model.

DEEP LEARNING SYSTEMS AND METHODS FOR PREDICTING IMPACT OF CARDHOLDER BEHAVIOR BASED ON PAYMENT EVENTS

Resumen de: WO2024025710A1

A system is configured to retrieve a set of customer raw transaction data, wherein the transactions are devoid of any target transactions of interest. An impact neural network model is applied to the transaction data using a "notTargef ' variable. The "notTargef ' variable indicates that the target transaction of interest is not included in the transaction data. The model predicts a first result based on the "notTargef' variable. The model is applied to the transaction data using an "isTargef ' variable. The "isTargef ' variable indicates that the target transaction of interest is included in the set of customer raw transaction data. The model predicts a second result based on the "isTargef ' variable. The system determines a difference between the second and first results. The difference is a predicted incremental impact on cardholder behavior. The system presents the predicted incremental impact on cardholder behavior to an issuer associated with the transaction data.

FORECASTING USING DIFFERENTIAL-BASED MACHINE-LEARNING ARCHITECTURE

Resumen de: WO2025108940A1

Conventional forecast models assume that the underlying data are stationary. In the case of non-stationary data, the conventional forecast model must be increased in complexity, or otherwise, must be frequently retrained. Embodiments are disclosed of a forecast model that accounts for non-stationary data, without increased complexity and without an increase in the frequency of retraining, by recasting the forecasting problem as a matter of tracking differentials in the data. In particular, variables (e.g., covariates) are input into differential determinations and/or intermediate machine-learning models to determine differences in past and forecasted values of a plurality of features. These differences are then input into a machine- learning model to predict a change in the value of the target, which is aggregated with a past value of the target to produce a forecasted value of the target.

GENERATING AND PROCESSING SIMULATED MEDICAL INFORMATION FOR PREDICTIVE MODELING

Resumen de: US2025174362A1

A system receives feature parameters, each identifying possible values for one of a set of features. The system receives outcomes corresponding to the feature parameters. The system generates a simulated patient population dataset with multiple simulated patient datasets, each simulated patient dataset associated with the outcomes and including feature values falling within the possible values identified by the feature parameters. The system may train a machine learning engine based on the simulated patient population dataset and optionally additional simulated patient population datasets. The machine learning engine generates predicted outcomes based on the training in response to queries identifying feature values.

LIFELONG MACHINE LEARNING (LML) MODEL FOR PATIENT SUBPOPULATION IDENTIFICATION USING REAL-WORLD HEALTHCARE DATA

Resumen de: US2025173563A1

A deep learning model implements continuous, lifelong machine learning (LML) based on a Bayesian neural network using a framework including wide, deep, and prior components that use available real-world healthcare data differently to improve prediction performance. The outputs from each component of the framework are combined to produce a final output that may be utilized as a prior structure when the deep learning model is refreshed with new data in a deep learning process. Lifelong learning is implemented by dynamically integrating present learning from the wide and deep learning components with past learning from models in the prior component into future predictions. The Bayesian deep neural network-based LML model increases accuracy in identifying patient profiles by continuously learning, as new data becomes available, without forgetting prior knowledge.

ARTIFICIAL INTELLIGENCE SYSTEM PROVIDING AUTOMATED DISTRIBUTED TRAINING OF MACHINE LEARNING MODELS

Resumen de: US2025173627A1

Multiple distinct control descriptors, each specifying an algorithm and values of one or more parameters of the algorithm, are created. A plurality of tuples, each indicating a respective record of a data set and a respective descriptor, are generated. The tuples are distributed among a plurality of compute resources such that the number of distinct descriptors indicated in the tuples received at a given resource is below a threshold. The algorithm is executed in accordance with the descriptors' parameters at individual compute resources.

Machine Learning Based Occupancy Forecasting

Resumen de: US2025173660A1

Embodiments determine a final occupancy prediction for a check-in date for a plurality of hotel rooms. Embodiments receive historical reservation data including a plurality of booking curves for the hotel rooms corresponding to a plurality of reservation windows, the historical reservation data including a plurality of features. Based on the historical reservation data, embodiments generate a first occupancy prediction for the check-in date using a first model and generate a second occupancy prediction for the check-in date using a second model. Embodiments determine a best performing model from at least the first model and the second model uses a corresponding occupancy prediction corresponding to the best performing model as the final occupancy prediction for the check-in date.

Monitoring Machine Learning Models

Resumen de: US2025173585A1

A computer-implemented method for monitoring machine learning models in a distributed setup includes obtaining model activity data relating to activity of the machine learning models in the distributed setup; analyzing the obtained model activity data; and, based on the analysis of the model activity data, outputting model management data for managing the activity of the machine learning models in the distributed setup.

SENSOR-BASED MACHINE LEARNING IN A HEALTH PREDICTION ENVIRONMENT

Resumen de: US2025174364A1

A machine learning prediction system can analyze a dataset of users with self-reported symptoms and associated data from a wearable device to impact measure the impact of an acute health condition (such as the flu) at the population level. The machine learning prediction system can train a machine learning model to recognize individual acute health condition patterns based on differences in user activity with respect to the characteristics of determined baseline periods. For example, per-individual normalized change with respect to baseline aggregated at the population level can be used to determine individual acute health condition patterns and predict the onset of certain acute health conditions using a trained machine learning model. In response to predictions, the machine learning prediction system can take interventions to manage the impact of a predicted acute health condition on an individual.

Methods And Systems For Quantifying Partitioning Of Agents In Vivo Based On Partitioning Of Agents In Vitro

Resumen de: US2025172540A1

Small molecule therapeutics can concentrate in distinct intracellular environments, some bounded by membranes, and others that may be formed by membrane-less biomolecular condensates. The chemical environments within biomolecular condensates have been proposed to differ from those outside these bodies, but the internal chemical environments of diverse condensates have yet to be explored. Here we use small molecule probes to demonstrate that condensates formed in vitro with the scaffold proteins of different biomolecular condensates harbor distinct chemical solvating properties. The chemical rules that govern selective partitioning in condensates, which we term condensate chemical grammar, can be ascertained by deep learning, allowing efficient prediction of the partitioning behavior of small molecules. The rules learned from in vitro condensates were adequate to predict the partitioning of small molecules into nucleolar condensates in living cells. Different biomolecular condensates harbor distinct chemical environments, that the chemical grammar of condensates can be ascertained by machine learning.

AI-CONTROLLED SENSOR NETWORK FOR THREAT MAPPING AND CHARACTERIZATION AND RISK ADJUSTED RESPONSE

Resumen de: US2025175456A1

A system and method for an AI-controlled sensor network for threat mapping and characterization. The system deploys a network of honeypots and sensors across various geographic locations and network segments, collecting and aggregating data on network traffic and potential threats. An AI orchestrator analyzes this data using advanced machine learning models, generating dynamic honeypot profiles and a comprehensive threat landscape. The system can adapt in real-time to emerging threats, optimize resource allocation, and provide actionable intelligence. By correlating data across multiple points, the system offers enhanced threat detection capabilities and proactive cybersecurity measures, surpassing traditional security information and event management (SIEM) tools.

SYSTEM AND METHOD FOR IDENTIFYING NATURAL ALTERNATIVES TO SYNTHETIC ADDITIVES IN FOODS

Resumen de: US2025173661A1

A method of modifying a food item to contain plant-based ingredients includes identifying plant-based substances to replace an ingredient of the food item. The plant-based substances are clustered, via a machine learning model, into a plurality of clusters according to an objective based on properties of the plant-based substances. The plant-based substances of a selected cluster are classified into a plurality of classes, via a machine learning classifier, based on the objective and the properties of the plant-based substances of the selected cluster. A score is determined for each plant-based substance of a selected class based on metrics. A plant-based substance is determined based on the score to produce a modified food item with the determined plant-based substance replacing the ingredient.

MACHINE LEARNING BASED METHODS OF ANALYSING DRUG-LIKE MOLECULES

Resumen de: US2025174314A1

There is provided a method for a machine learning based method of analysing drug-like molecules by representing the molecular quantum states of each drug-like molecule as a quantum graph, and then feeding that quantum graph as an input to a machine learning system.

AUTOMATIC REVISIONS TO DOCUMENT CLAUSES BASED ON CLAUSE TYPE

Resumen de: US2025173380A1

A document management system can include an artificial intelligence-based document manager that can perform one or more predictive operations based on characteristics of a user, a document, a user account, or historical document activity. For instance, the document management system can apply a machine-learning model to determine how long an expiring agreement document is likely to take to renegotiate and can prompt a user to begin the renegotiation process in advance. The document management system can detect a change to language in a particular clause type and can prompt a user to update other documents that include the clause type to include the change. The document management system can determine a type of a document being worked on and can identify one or more actions that a corresponding user may want to take using a machine-learning model trained on similar documents and similar users.

FORECASTING USING DIFFERENTIAL-BASED MACHINE-LEARNING ARCHITECTURE

Resumen de: EP4560561A1

Conventional forecast models assume that the underlying data are stationary. In the case of non-stationary data, the conventional forecast model must be increased in complexity, or otherwise, must be frequently retrained. Embodiments are disclosed of a forecast model that accounts for non-stationary data, without increased complexity and without an increase in the frequency of retraining, by recasting the forecasting problem as a matter of tracking differentials in the data. In particular, variables (e.g., covariates) are input into differential determinations and/or intermediate machine-learning models to determine differences in past and forecasted values of a plurality of features. These differences are then input into a machine-learning model to predict a change in the value of the target, which is aggregated with a past value of the target to produce a forecasted value of the target.

SECURITY-RELATED EVENT ANOMALY DETECTION

Resumen de: US2025168179A1

The technology relates to machine responses to anomalies detected using machine learning based anomaly detection. In particular, to receiving evaluations of production events, prepared using activity models constructed on per-tenant and per-user basis using an online streaming machine learner that transforms an unsupervised learning problem into a supervised learning problem by fixing a target label and learning a regressor without a constant or intercept. Further, to responding to detected anomalies in near real-time streams of security-related events of tenants, the anomalies detected by transforming the events in categorized features and requiring a loss function analyzer to correlate, essentially through an origin, the categorized features with a target feature artificially labeled as a constant. An anomaly score received for a production event is determined based on calculated likelihood coefficients of categorized feature-value pairs and a prevalencist probability value of the production event comprising the coded features-value pairs.

Closed-Loop Intelligence

Resumen de: US2025165439A1

Methods, computer systems, and computer-storage medium are provided for providing closed-loop intelligence. A selection of data is received, at a cloud service, from a database comprising data from a plurality of sources in a Fast Healthcare Interoperability Resources (FHIR) format to build a data model. After a feature vector corresponding to the data model is extracted, a selection of an algorithm for a machine learning model to apply to the data model is received. A portion of the selection of data is utilized for training data and test data and the machine learning model is applied to the training data. Once the model is trained, the trained machine learning model can be saved at the cloud service, where it may be accessed by others.

Closed-Loop Intelligence

Resumen de: US2025165438A1

Methods, computer systems, and computer-storage medium are provided for providing closed-loop intelligence. A selection of data is received, at a cloud service, from a database comprising data from a plurality of sources in a Fast Healthcare Interoperability Resources (FHIR) format to build a data model. After a feature vector corresponding to the data model is extracted, a selection of an algorithm for a machine learning model to apply to the data model is received. A portion of the selection of data is utilized for training data and test data and the machine learning model is applied to the training data. Once the model is trained, the trained machine learning model can be saved at the cloud service, where it may be accessed by others.

INTERACTIVE REAL-TIME VIDEO SEARCH BASED ON KNOWLEDGE GRAPH

Resumen de: US2025165529A1

A method, computer system, and computer program product are provided for real-time video searching based on augmented knowledge graphs that are generated using machine learning models. Multimedia data is obtained comprising an image portion and an audio portion, and a user query with respect to the multimedia data is obtained. A knowledge graph of the multimedia data is generated using one or more machine learning models based on the image portion and the audio portion, wherein the knowledge graph includes a plurality of entities and relationships between entities. An augmented knowledge graph is generated, wherein the augmented knowledge graph augments the knowledge graph with additional entities and additional relationships between the additional entities using additional data that is obtained from a source external to the multimedia data. A response to the user query is provided based on the augmented knowledge graph.

STORING AND OBTAINING ATTRIBUTE DATA OF ATTRIBUTES OF MACHINE LEARNING MODELS

Resumen de: US2025165816A1

In some implementations, a controller may receive a request for an inference. The controller may determine, based on the received request for the inference, a first inference model of a plurality of inference models, to generate the inference. The controller may obtain, from a memory associated with an inference cache, first attribute data regarding first attributes of the first inference model. A location of the first attribute data, in the memory, may be determined using the inference cache. The attributes may include weights associated with the first inference model, biases associated with the first inference model, and a structure of the first inference model. The controller may utilize the first attribute data to generate the inference based on the request.

GENERATING AN EMBEDDING SPACE FROM A TRAINING DATA SET FOR A MACHINE LEARNING ALGORITHM

Resumen de: US2025165865A1

A system for and a method of generating an embedding space from a training data set for a machine learning algorithm. The method includes receiving, through an input of a computer system, an input operating data point by one or more layers of a machine learning algorithm implemented on the computer system, generating an embedding space using a training data set, the embedding space including a plurality of inducing data points, each inducing data point corresponding to a cluster of training data points, comparing the input operating data point against the plurality of inducing data points, determining whether a sufficient set of supporting historical data points exists in the plurality of training data points within a fixed distance from the input operating data point, and generating an alert based on the determining to enable a user or an electronic controller to take appropriate action, and outputting the alert.

SYSTEMS AND METHODS FOR ITERATIVE FEATURE SELECTION FOR MACHINE LEARNING MODELS

Nº publicación: US2025165848A1 22/05/2025

Solicitante:

CAPITAL ONE SERVICES LLC [US]
Capital One Services, LLC

Resumen de: US2025165848A1

Systems and methods for selecting machine learning features using iterative batch feature reduction. In some aspects, the system trains a plurality of candidate models based on a plurality of feature groups split from a first set of features. Each candidate model takes as input a feature group of no more than a first threshold number of features. For each candidate model in the plurality of candidate models, the system processes the candidate model to extract an explainability vector. Based on the explainability vector, the system selects a second threshold number of features from the feature group to generate a slim feature group. The system trains a slim candidate model which takes as input the slim feature group. The system generates a second set of features by combining features from a plurality of slim candidate models.