Alerta

METHODS AND APPARATUSES FOR EXPLAINABLE AI-BASED NETWORK-CONTROLLED SYSTEM OPTIMIZATION

Absstract of: US20260230843A1

0000 A wireless transmit/receive unit (WTRU) may be configured to determine one or more explainability artificial intelligence (XAI) capabilities of the WTRU. The WTRU may receive XAI configuration information from a network that indicates one or more of an explainability mode, an explainability domain type, a configuration of XAI outcome storage, one or more triggers associated with an XAI report, or an XAI report configuration. The WTRU may determine one or more XAI parameters associated with an XAI model. The one or more XAI parameters may indicate features that have a greatest impact on predictions made by an artificial intelligence (AI) or machine learning (ML) inference model. The WTRU may determine to trigger an XAI report based on one or more outcomes of the XAI model. The WTRU may send the XAI report to the network.

METHODS TO SUPPORT AIML FALLBACK OPERATION

Absstract of: US20260230845A1

0000 A method may be implemented by a wireless transmit/receive unit (WTRU), including receiving information indicating first applicability conditions and a first configuration for a first artificial intelligence or machine learning (AIML) model category and second applicability conditions and a second configuration for a second AIML model category. It may be determined whether the first AIML model category is applicable to a cell based on the first applicability conditions and first configuration, and whether the second AIML model category is applicable based on the second applicability conditions and second configuration. A report may be sent indicating applicability of the first AIML model category and second AIML category to the cell. It may be determined that the first AIML model category is no longer applicable based on the first configuration. An indication that indicates that the first AIML model category is no longer applicable to the cell may be sent.

Model Explainability Based on an Ensemble Approach

Absstract of: US20260228577A1

0000 Methods are provided for generation of natural-language explanations for machine learning model outputs that exhibit reduced hallucinations and higher accuracy while incurring lower computational costs and enhancing privacy. These methods include determining, for the textual and categorical features of an input, respective importance values regarding the degree to which the overall model output was influenced by the input features. These values, along with the input features themselves and the overall model output, are then applied to a natural language model to generate an explanation of the overall model output with respect to the input features. The importance values can be used to prune input features that were least important to the overall model output, further reducing hallucinations, increasing accuracy, and reducing overall computational cost by allowing models with smaller input widths or otherwise less computationally expensive natural language models to be used.

MACHINE LEARNING FEATURE RECOMMENDATION

Absstract of: US20260228609A1

A specification of a desired target field for machine learning prediction and one or more tables storing machine learning training data are received. Within the one or more tables, eligible machine learning features for building a machine learning model to perform a prediction for the target field are identified. The eligible machine learning features are evaluated using a pipeline of different evaluations to successively filter out one or more of the eligible machine learning features to identify a set of recommended machine learning features among the eligible machine learning features. The set of recommended machine learning features is provided for use in building the machine learning model.

SYSTEMS AND ASSOCIATED METHODS FOR FEATURE RECOGNITION

Absstract of: US20260228377A1

0000 A system for feature recognition of design geometry according to an implementation of the present disclosure may include one or more processors coupled to memory. The one or more processors may be collectively operable to execute a mapping environment. The mapping environment may be operable to access at least one component design including a plurality of geometric features. The mapping environment may be operable to determine values for geometric attributes associated with the respective geometric features of the component design. The mapping environment may be operable to assign, using a machine learning classifier model, predefined names to the respective geometric features based on the values of the geometric attributes. The mapping environment may be operable to assign values for one or more parameters to the respective geometric features according to the assigned predefined names. A method for feature recognition of design geometry is also disclosed.

DATA HANDLING OF HOMOMORPHIC ENCRYPTED DATA USED FOR MACHINE LEARNING MODEL TRAINING IN O-RAN

Absstract of: WO2026165011A1

Embodiments of the disclosure describe a method for data handling of homomorphic encrypted data used for machine learning model training in an Open Radio Access Network (O-RAN). The method includes generating, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling. The method further includes transmitting, by the data source entity, the generated data to a model training entity. The generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model. The trained ML model is transferred from the model training entity to a model inference entity. The method further includes transmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

SYSTEMS AND METHODS FOR CONTROLLING AUTONOMOUS ROBOTS USING MACHINE LEARNING BASED PLANNING

Absstract of: WO2026162001A1

Systems and methods for controlling autonomous robots using machine learning based planning are proposed that utilize an improved computational model that is configured to infers a small distribution over possible programmatic cognitive maps conditioned on human prior knowledge of the world, and uses this distribution to generate resource-efficient plans, using a trained Large Language Model as an embedding of human priors. Fragment based computation approaches to allow for computational reuse are also proposed. By using large language model, the system can generate programmatic map representation and fragment-based planning.

AUTOMATED PERFORMANCE MONITORING AND RETRAINING OF MACHINE LEARNING MODELS

Absstract of: US20260228603A1

0000 Disclosed are systems and methods that automatically monitor the performance of machine learning models operating in a production system and that improve them through automated retraining to meet configured performance standards.

SCALABLE MACHINE LEARNING PLATFORM WITH REAL-TIME MODEL SERVING

Absstract of: WO2026165123A1

A system and method for machine learning platform management includes a unified architecture for developing and deploying machine learning models at scale. The platform integrates feature generation, model training, and inference services through a centralized interface. The system processes source data through a feature platform to generate training datasets and real- time features. A multi-stage training pipeline enables automated model experimentation through configurable workflows combining core frameworks and user modeling code. The platform implements specialized inference services optimized for high-throughput ranking and recommendation use cases, with distributed feature stores and local caching for efficient feature serving. A comprehensive monitoring system tracks model performance, feature distributions, and prediction quality through automated anomaly detection. The platform enables rapid experimentation while maintaining production reliability through automated deployment orchestration, optimized inference engines, and continuous feedback loops for model improvement.

MULTI-TASK MACHINE LEARNING MODELS FOR DIGITAL PATHOLOGY

Absstract of: WO2026165288A2

Techniques for using multi-task machine learning models for digital pathology are described herein. In an example, a system accesses an image of a biological sample of a patient having a medical condition. The system can process one or more embeddings using two or more models to obtain a primary continuous variable associated with an outcome for the patient and one or more auxiliary properties of the biological sample.

Systems and Methods for the Prediction of Post-Operative Cognitive Decline Using Blood-Based Inflammatory Biomarkers

Absstract of: US20260229316A1

Embodiments herein describe systems and methods to generate a risk score for an individual to develop postoperative neurocognitive disorder (POND). Various embodiments obtain multi-omics data from an individual, such as genomics, transcriptomics, and proteomics. In certain embodiments, a machine learning algorithm is used to generate the risk score based on the multi-omics data. In further embodiments, clinical data is further used in the determination of the risk score.

METHOD AND SYSTEM FOR GENERATION OF AT LEAST ONE OUTPUT ANALYTICS FOR A PROMOTION

Absstract of: US20260228773A1

0000 There is provided a method and system for generating an output analytic for a promotion. The method includes training and instantiating a machine learning model comprising at least a Random Forest model, with a selection training set, the selection training set comprising the historical data and the one or more input parameters; selecting, by the processor, using the machine learning model a configuration and a layout for the one or more products on the promotional materials; outputting, by the processor, the promotional materials based on the selection of the configuration and layout.

Deterministic Inference for Machine Learning Models with Variable Behavior

Absstract of: US20260228561A1

0000 Provided are systems and methods that enable deterministic inference for machine learning models with variable behavior. In particular, the present disclosure relates to a system in which a machine-learned model has a variable processing portion that is configured to variably apply one or more of a plurality of different processing operations when processing an input. According to an aspect of the present disclosure, one or more seed values can be used to deterministically control which of the plurality of different processing operations are applied by the machine-learned model when processing a given set of input data.

SYSTEMS AND METHODS FOR IDENTIFYING BREATHING COMFORT

Absstract of: US20260224834A1

A method for classifying a user of a respiratory pressure therapy (RPT) device is disclosed, the method comprising: identifying that the user has commenced a therapy session with the RPT device; receiving first input data associated with the user and the therapy session; training a machine learning model based on a plurality of pre-classified users and the first input data to determine an association between a user and a plurality of classifications; classifying the user into a classification of the plurality of classifications; and in response to classifying the user into the classification, causing a user device and/or the RPT device to execute an action. The first input data may include selections from the user in response to a survey prompt.

ETHICAL AND SAFE ARTIFICIAL GENERAL INTELLIGENCE (AGI)

Absstract of: US20260228491A1

0000 The existential question for Artificial General Intelligence (AGI) is whether the values of AGI will align with human values. Solving “the Alignment Problem” is critical. Get it right, and we unlock trillions of dollars of productivity and huge benefits for humanity. Get it wrong and humanity goes extinct. This invention shows how to design ethical and safe AGI that solves the Alignment Problem. The invention includes scalable ethics and safety features as well as several learning, training, tuning, and customization methods that go beyond the standard techniques of machine learning such as Transformers and Deep Learning techniques. The AGI is implemented using either external problem solvers connected on a network or internal AI agents collaborating within a single computerized system. Detailed implementation examples—revealing technological and economic synergy with Meta, Amazon, Google, DeepMind, YouTube, TikTok, Microsoft, OpenAI, Twitter/X, Tesla, Nvidia, Tencent, Apple, and Anthropic—are described.

TRAINING A MACHINE LEARNING SYSTEM FOR TRANSACTION DATA PROCESSING

Absstract of: EP4787244A2

A method of training a supervised machine learning system to detect anomalies within transaction data is described. The method includes obtaining a training set of data samples; assigning a label indicating an absence of an anomaly to unlabelled data samples in the training set; partitioning the data of the data samples in the training set into two feature sets, a first feature set representing observable features and a second feature set representing context features; generating synthetic data samples by combining features from the two feature sets that respectively relate to two different uniquely identifiable entities; assigning a label indicating a presence of an anomaly to the synthetic data samples; augmenting the training set with the synthetic data samples; and training a supervised machine learning system with the augmented training set and the assigned labels.

METHOD AND SYSTEM FOR INTERPRETING COMMON INTERACTION UTILITY AMONGST MULTIPLE BLACKBOX ARTIFICIAL INTELLIGENCE MODELS

Absstract of: EP4787251A1

0001 The present application relates to the technical field of machine learning and discloses a method and system for interpreting common interaction effect among multiple black-box artificial intelligence models. The method and system can automatically analyze common interactions modeled among multiple different artificial intelligence models, and can also analyze common interactions modeled by the same model when the input is subjected to different perturbations. The implementation of the method and system comprises the following steps: providing an input sample; using multiple black-box models to perform prediction on the same data, or using a single black-box model to perform prediction on different data, to obtain multiple sets of prediction results; based on the multiple sets of prediction results, modeling the interactions among input units of the sample, computing the interaction strengths of combinations formed among input units, and expressing the output of each set of models as "AND interaction effect" and "OR interaction effect" among input unit combinations; and learning common "AND interaction effect" and "OR interaction effect" shared among different artificial intelligence models.

DETERMINATION OF THE ESTIMATED LENS POSITION OF AN INTRAOCULAR LENS USING THREE-DIMENSIONAL ANTERIOR SEGMENT GEOMETRY AND MACHINE LEARNING

Absstract of: WO2025106772A1

A method for determining the relationship between measured variables of the anterior segment in a patient's eye pre-operatively and the post-operative position, and optionally tilt, of the implanted intraocular lens (IOL) is described, based on quantitative optical coherence tomography imaging of the eye of a patient and a machine learning architecture to determine the best regression model. Formulas are further described obtained using such method to determine the Estimated Lens Position (ELP), and optionally tilt, from measured variables in a patient pre-operatively to incorporate in IOL power calculation formulas or ray-tracing based IOL power selection.

ENSEMBLE MACHINE LEARNING FOR TIME SERIES DATA WITH GAPS

Absstract of: WO2025071598A1

Ensemble machine learning can be used to make predictions based on time series data with gaps. Multiple models are trained on different (overlapping) sets or portions of the available time series data, and the predictions from the different models are aggregated to generate predictions. The models can include one model trained on all of the time series data and a second model trained using just the data points that immediately follow the gaps. Models in an ensemble can also include models that use all features of the data points and models that use only a subset of features of the data points.

REAL-TIME FINANCIAL INTELLIGENCE

Absstract of: US20260220709A1

0000 At least one example describes mechanisms for real-time financial intelligence (RFI) that leverage one or more machine-learning (ML) techniques to process financial data. In at least one example, RFI may dynamically generate dimensions and their associated contextual information within the financial data to enable dimensional intelligence. In at least one example, dimensional intelligence may include performing intelligent data analytics across the dimensions, enabling generation of meaningful insights, efficient reporting, identifying patterns, and supporting decision-making. In at least one example, RFI may offer an interactive graphical user interface-financial intelligent chat (Fine-Chat), enabling users to input financial queries in a natural language format. In at least one example, via these financial queries, Fine-Chat may perform dimensional intelligence, dimensional analysis, intelligent data analytics and generates real-time financial reports, financial recommendations. Fine-Chat may generate a response to a financial query including at least one result in a natural language format.

AN IMPROVED METHOD OF, AND APPARATUS FOR, MEDICAL DATA OUTCOME ANALYSIS

Absstract of: WO2026158807A1

An Improved Method of, and Apparatus for, Medical Data Outcome Analysis There is provided a computer-implemented method for determining an uncertainty metric for a medical data abstraction process utilising a federated machine learning model executed on a computing node and having a plurality of personal abstraction models (PAMs) associated therewith, wherein each PAM comprises an instance of the federated machine learning model local to and executed on a client computing node and trained using a local training dataset stored on the respective client node, wherein the method is executed by at least one hardware processor and comprises the steps of: inputting medical data relating to a patient medical record into the federated machine learning model to generate an outcome classification prediction for one or more medical parameters; inputting the medical data relating to the patient medical record into each trained PAM to generate an outcome classification prediction for the one or more medical parameters for each trained PAM; utilising the determined outcome classification predictions from each trained PAM in an ensemble voting process to determine an uncertainty metric of the outcome classification prediction in step a); d) determining whether the uncertainty metric meets one or more predefined criteria and, if so, accepting the determined outcome classification prediction and abstracting the patient medical record in an abstraction database.

SYSTEMS, METHODS AND TESTING APPARATUS FOR ASSESSING PLATELET SAMPLES USING MACHINE LEARNING FOR CANCER ESTIMATION

Absstract of: WO2026156436A1

An artificial intelligence-based method for detecting a disease comprises a meta- classification model comprising a plurality of base classification models each generating a respective output and wherein each of such outputs is input to a meta-model, the meta-model comprising an artificial neural network. Variants are described in respect of the meta-model stacked ensemble pipeline, dimensionality / complexity control, and specific architectures designed for tracking disease trajectory through identifying gene pathway crosstalk. The method is particularly suitable for detecting cancer in blood samples comprising tumor educated platelets.

METHODS AND SYSTEMS FOR EMPOWERING MACHINE UNLEARNING VIA CONTRASTIVE LEARNING

Absstract of: WO2026157104A1

A method for unlearning data in a machine learning system involves generating a first unlearned model by performing a first unlearning iteration on a global model, based on a request from a target client. The server transmits the unlearned model to the client, which verifies, using a metric, whether further unlearning is required. If needed, the target client sends a notification to a network controller, which selects an unlearning function from a function pool and transmits it to the server. The server performs a second unlearning iteration using the selected function, generating a second unlearned model. The second model is transmitted to the target client for verification. Upon confirmation that the model meets unlearning requirements, the server broadcasts the refined model to non-target clients. The method ensures effective data removal while maintaining model integrity for continued use across the system.

METHOD TO IDENTIFY AND MODIFY NWDAF DEPENDENCIES

Absstract of: US20260222304A1

0000 A computer-implemented method performed by a computing device including a NWDAF having a function to identify and modify dependencies between NWDAFs and at least a first NWDAF consumer. The method includes identifying a relationship between at least the first NWDAF consumer from a plurality of NWDAF consumers that respectively subscribe to the NWDAFs that have respective machine learning, ML, models that have an overlap between a first input feature and respectively have different outputs. At least the first NWDAF consumer uses a first output from the NWDAFs to output an action from the first NWDAF consumer. The method further includes determining whether a second input feature to (i) a respective ML model of the NWDAFs or (ii) a new ML model of an unsubscribed NWDAF can make a positive contribution to the prediction of the output of at least the first NWDAF consumer.

CLEANING A MACHINE LEARNING DATASET USING CORESETS

Nº publicación: US20260220535A1 30/07/2026

Applicant:

DATAHEROES LTD [IL]
DATAHEROES, LTD.

Absstract of: US20260220535A1

A computerized method of cleaning a data set comprises: (a) providing the data set, configured for training of a machine learning model, and comprising data instances. Each instance comprises a label; (b) selecting a number of data instances from the set based on importance-related criteria, indicative of corresponding high importance values associated with the data instances. This generates high importance set(s), giving rise to a set of selected data instances having a high probability of high importance values, as compared to a second probability associated with selecting data instances from the data set. The selected set facilitates determining, for each data instance of, whether it requires an action. This facilitates performing the action, which brings about increased quality data set(s), which facilitate quicker and/or higher-accuracy training of the machine learning model, as compared to a second training of the machine learning model performed utilizing the data set.