Alerta

CPU USAGE RATE PREDICTION METHOD AND APPARATUS, AND COMPUTER DEVICE, READABLE STORAGE MEDIUM AND STORAGE PRODUCT

Resumen de: WO2026118833A1

The present application relates to a CPU usage rate prediction method and apparatus, and a computer device and a readable storage medium. The method comprises: during the running of a target software application, for each node in a node cluster in which the target software application is deployed, acquiring, in real time, baseline CPU usage rates of a central processing unit (CPU) of the node at a preset number of time points prior to the current time point; inputting the baseline CPU usage rates into a self-attention-mechanism-based deep learning network model, a convolutional neural network model and a recurrent neural network model, so as to obtain a first feature vector, a second feature vector and a third feature vector; and fusing the first feature vector, the second feature vector and the third feature vector, and on the basis of a fused feature vector, obtaining the current CPU usage rate of the node at the current time point.

SIDELOBE SUPPRESSION METHOD AND APPARATUS FOR RADAR PLANE IMAGING, ELECTRONIC DEVICE AND COMPUTER PROGRAM PRODUCT

Resumen de: WO2026118410A1

Provided in the present application are a sidelobe suppression method and apparatus for radar plane imaging, an electronic device and a computer program product. The method comprises: acquiring a constructed training data set, the training data set comprising radar plane image samples containing sidelobes and corresponding radar plane image samples without sidelobes or with sidelobes suppressed; on the basis of the training data set, training a constructed adaptive neural network model to obtain an adaptive sidelobe suppression model; acquiring a radar plane image to be processed; and, on the basis of the adaptive sidelobe suppression model, performing adaptive sidelobe suppression on the radar plane image to be processed. The solution of the present application essentially achieves sidelobe suppression for radars by using a signal processing means, thus enabling direct processing of radar images in a completely deployed radar system without changing hardware facilities, and allowing for dynamically adjusting the sidelobe suppression strategy according to different task requirements and scenarios; thus, the more flexible and post-deployment adjustable solution can enhance the adaptability and maintainability of radar systems.

METHOD, APPARATUS, AND MEDIUM FOR VIDEO PROCESSING

Resumen de: US20260164047A1

Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. The method comprises: obtaining a neural network (NN) model for processing a video, the NN model comprising at least one basic block, wherein a basic block comprises: a plurality of branches for parallel processing an input of the basic block, a branch comprising at least one convolutional layer and at least one activation layer, and a plurality of layers for serial processing a combination of outputs of the plurality of branch, the plurality of layers comprising at least one convolutional layer and at least one activation layer; and performing, according to the NN model, a conversion between a current video block of the video and a bitstream of the video.

MULTI-VIEW FUSION METHOD FOR FORECASTING POWER OF WIND TURBINES IN OFFSHORE WIND FARM

Resumen de: US20260160234A1

Provided is a multi-view fusion method for forecasting power of wind turbine group in an offshore wind farm including obtaining a feature vector, constructing a power relationship matrix and a graph matrix based on data of wind turbine group and a geographical position of each wind turbine, constructing a spatial graph embedding module, to embed the graph matrix into node spatial information and inter-graph node information, inputting an embedding information matrix to a cross-fusion convolution module, constructing a Chebyshev graph convolutional neural network to process the feature vector and a multi-view topology matrix, to enable the feature vector of each wind turbine to obtain effective weights of other wind turbine group, and finally screening, by a multi-timing gating module, a time sequence feature. In the method, structural features of the wind turbine group can be fully captured, and dynamic features of the wind turbine group can be fully captured.

GENERATIVE NEURAL NETWORKS WITH INVISIBLE TOKENS

Resumen de: US20260161713A1

Methods, systems, and apparatuses, including computer programs encoded on computer storage media, for processing a network input using a generative neural network to generate an output sequence of output tokens. The system selects each output token from a vocabulary of tokens that includes a plurality of visible tokens and one or more pairs of invisible tokens. The system processes the output sequence of output tokens to generate a final output sequence by removing, from the output sequence, the beginning invisible token, the end invisible token, and each visible token that is between the beginning invisible token and the end invisible token. The system then provides the final output sequence in response to the network input.

RECOMMENDATION SYSTEM AND RECOMMENDATION METHOD FOR INFERRING EDGE INFORMATION IN A HYPERBOLIC SPACE AND PERFORMING USER-SPECIFIC RECOMMENDATION

Resumen de: US20260161740A1

The present invention relates to a recommendation method for performing personalized recommendation by inferring edge information in a hyperbolic space and, more particularly, to a technique in which a recommendation system server collects review data corresponding to users and items, generates review document vectors and maps the review document vectors into the hyperbolic space, computes similarity scores based on hyperbolic distances, infers edge existence probabilities through a link prediction neural network, and generates a final recommended item list for each user by reflecting domain-shared features across multiple domains.

IMAGE PROCESSING APPARATUS AND OPERATING METHOD THEREOF

Resumen de: US20260162420A1

An image processing apparatus includes memory storing one or more instructions and at least one processor. When executed, the instructions cause the apparatus to obtain a power consumption reduction request and, in response, obtain, from pre-stored profiling data of a first neural network model, a threshold value for converting one or more parameters of the first neural network model to zero. The profiling data includes information indicating the threshold value, performance information for a second neural network model generated by converting the one or more parameters of the first neural network model to zero based on the threshold value, and power consumption reduction estimation information for the second neural network model. The instructions further cause the apparatus to obtain an output image from the second neural network model by processing an input image through the second neural network model in which the one or more parameters are converted to zero.

PEDESTRIAN BEHAVIOR PREDICTION WITH 3D HUMAN KEYPOINTS

Resumen de: US20260159135A1

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for agent behavior prediction using keypoint data. One of the methods includes obtaining data characterizing a scene in an environment, the data comprising: (i) context data comprising data characterizing historical trajectories of a plurality of agents up to the current time point; and (ii) keypoint data for a target agent; processing the context data using a context data encoder neural network to generate a context embedding for the target agent; processing the keypoint data using a keypoint encoder neural network to generate a keypoint embedding for the target agent; generating a combined embedding for the target agent from the context embedding and the keypoint embedding; and processing the combined embedding using a decoder neural network to generate a behavior prediction output for the target agent that characterizes predicted behavior of the target agent after the current time point.

Cross-Modal Contrastive Learning for Text-to-Image Generation based on Machine Learning Models

Resumen de: US20260162320A1

A computer-implemented method includes receiving, by a computing device, a particular textual description of a scene. The method also includes applying a neural network for text-to-image generation to generate an output image rendition of the scene, the neural network having been trained to cause two image renditions associated with a same textual description to attract each other and two image renditions associated with different textual descriptions to repel each other based on mutual information between a plurality of corresponding pairs, wherein the plurality of corresponding pairs comprise an image-to-image pair and a text-to-image pair. The method further includes predicting the output image rendition of the scene.

FIXATION TOLERANT OBJECT RECOGNITION WITH A LOGARITHMICALLY-COMPRESSED VISUAL FILED

Resumen de: US20260162418A1

0000 Systems and methods are described herein that utilize neural networks to learn and implement convolutional filters that can be used over a logarithmically-compressed image. These filters are tolerant to changes in the relative visual fixation of an image, that is, a change of origin in log-polar coordinates. Deep networks, such as those that use multilayered neural networks, may be configured to implement the proposed method for filter learning and to take advantage of the exponential savings associated with a logarithmically-compressed image space with minimal sacrifice of fixation invariance.

CONTROL SYSTEM WITH NEURAL NETWORK PREDICTOR INCORPORATING FIRST PRINCIPLES

Resumen de: US20260161145A1

A predictive control system for a plant includes a predictor model trainer and a predictive controller. The predictor model trainer is configured to train a predictor model using a loss function including (i) a first error loss term based on an error between predicted values of controlled variables (CVs) generated by the predictor model and historical values of the CVs in historical state data and (ii) a second error loss term based on the predicted values of the CVs and physical relationships involving the CVs. The predictive controller is configured to control operation of the plant using the trained predictor model.

Neural Networks based Multimodal Transformer for Multi-Task User Interface Modeling

Resumen de: US20260162416A1

0000 A method includes receiving, via a computing device, a screenshot of a display provided by a graphical user interface of the computing device. The method also includes generating, by an image-structure transformer of a neural network, a representation by fusing a first embedding based on the screenshot and a second embedding based on a layout of virtual objects in the screenshot. The method additionally includes predicting, by the neural network and based on the generated representation, a modeling task output associated with the graphical user interface. The method further includes providing, by the computing device, the predicted modeling task output.

NEURAL NETWORK POWER MANAGEMENT IN A MULTI-GPU SYSTEM

Resumen de: US20260161215A1

Systems, apparatuses, and methods for managing power consumption for a neural network implemented on multiple graphics processing units (GPUs) are disclosed. A computing system includes a plurality of GPUs implementing a neural network. In one implementation, the plurality of GPUs draw power from a common power supply. To prevent the power consumption of the system from exceeding a power limit for long durations, the GPUs coordinate the scheduling of tasks of the neural network. At least one or more first GPUs schedule their computation tasks so as not to overlap with the computation tasks of one or more second GPUs. In this way, the system spends less time consuming power in excess of a power limit, allowing the neural network to be implemented in a more power efficient manner.

3 A METHOD AND AN APPARTUS FOR GENERATING 3D INFORMATION OF AN OBJECT USING A NEURAL NETWORK MODEL

Resumen de: KR20260086642A

본 개시는 뉴럴 네트워크 모델을 이용하여 객체의 3차원 정보를 생성하는 방법 및 장치에 관한 것이다. 일 실시예에 따르면, 뉴럴 네트워크 모델을 이용하여 객체의 3차원 정보를 생성하는 방법은 선박 환경을 촬영한 적어도 하나의 제1 이미지를 이용하여 뉴럴 네트워크 모델을 학습시키는 단계; 상기 학습된 뉴럴 네트워크 모델에 제2 이미지를 입력하는 단계; 및 상기 뉴럴 네트워크 모델로부터 상기 제2 이미지에 포함된 객체의 3차원 정보를 획득하는 단계;를 포함할 수 있다.

A method for fusing at least two output signals of at least two sensor devices of a motor vehicle, a corresponding computer program product, a corresponding non-transitory computer-storage medium, as well as corresponding electronic computing device

Resumen de: GB2702296A

A method for fusing at least two output signals (22, 24) of at least two sensor devices (14, 16) of a motor vehicle (10) by an electronic computing device (12) of the motor vehicle, comprising the steps of: receiving at least a first output signal (22) of at least a first sensor device (14) of the at least two sensor devices by the electronic computing device; receiving at least a second output signal (24) of at least a second sensor device (16) of the at least two sensor devices by the electronic computing device; providing a deep neural network (30) for determining a time offset between the at least two received output signals by the electronic computing device; determining the time offset by the deep neural network; fusing the at least two output signals depending on the determined time offset by the electronic computing device. Furthermore, the present invention relates to a computer program product, a non-transitory readable computer-storage medium, as well as an electronic computing device. Fig. 1

Processor

Resumen de: GB2702268A

A processor has a neural processing unit, or neural engine 700, which in turn has a local storage 738; a handling unit 720 generates invocation data to load a block of a tensor into the local storage from a storage of the processor; the tensor has a first number of dimensions, and the block of the tensor has a size of one in one of said dimensions such that the block consists of tensor elements arrayed in a second number of dimensions, the second number of dimensions is fewer than the first; a storage access controller receives the generated invocation data and identifies a position of the block within the tensor, then loads data corresponding to the identified block of the tensor into the local storage; one or more execution sub-unit of the neural processing unit performs an operation on the block. Preferably the tensor represents a feature map of a neural network. The storage access controller may be made up of an input reader 724, output writer 726, and DMA unit 728. The storage of the processor may be made up of an L1 cache (656a, 656b; figure 1b) or an L2 cache (660). A neural execution description (NED) may include an external storage stride. figure 2

APPARATUS AND METHOD FOR ASSISTING CYBER THREAT DETECTION AND RESPONSE BASED ON NEURAL NETWORK PROCESSING UNIT

Resumen de: KR20260082451A

본 개시는 신경망 처리 장치 기반의 사이버 위협 탐지 및 대응을 위한 어시스턴트 장치 및 방법에 관한 것으로, 특히, 장치는, 외부 장치와 통신을 수행하는 통신모듈; 신경망 처리 장치 기반의 사이버 위협 탐지 및 대응을 위한 어시스턴트 동작을 수행하기 위한 적어도 하나의 프로세스가 저장된 메모리; 및 상기 프로세스에 따라 상기 어시스턴트 동작을 수행하는 프로세서를 포함하며, 상기 프로세서는, 페이로드를 데이터로서 수집하여 전처리하고, 전처리된 데이터를 사전 학습된 인공지능 기반 탐지 모델로 입력하고, 상기 탐지 모델로부터 사전 설정된 판단기준에 기초한 공격 유무에 대한 판단 결과가 출력되면, 상기 판단 결과를 언어 모델에 입력하고, 상기 언어 모델로부터 상기 판단 결과에 기초한 답변이 출력되면, 상기 답변을 사용자에게 제공하도록 구성될 수 있다. '과학기술정보통신부KISA의 "2024년 AI 보안 제품 및 서비스 사업화 지원사업" 산출물 또는 지원을 받아서 상용화된 제품·서비스'

NEURAL NETWORK TRAINING USING RESAMPLED IMAGE DATA

Resumen de: US20260154378A1

0000 Apparatuses, systems, and techniques to modify a set of training data used for machine learning. In at least one embodiment, a set of images used for training a machine learning system is resampled by augmenting the set of images with additional images of under represented object types extracted from portions of existing training images in the set.

TRAINING A NEURAL NETWORK TO SIMULTANEOUSLY ASCERTAIN SEMANTIC INFORMATON AND DEPTH INFORMATION

Resumen de: US20260154979A1

0000 A method for training an image processing neural network. The method includes: providing a set of training images; feeding each training image to a first trained neural network, which assigns semantic information to pixels, other image portions, and/or image features of an input image; feeding each training image to a second trained neural network, which assigns depth information to pixels, other image portions, and/or image features of an input image; fusing the semantic information and depth information to form a target map, which assigns semantic information to locations in three-dimensional space; processing, using the image processing neural network to be trained, each training image to form a map, which assigns semantic information to locations in three-dimensional space; checking, using a cost function, to what extent the map thus obtained is in line with the target map; optimizing parameters that characterize the behavior of the image processing neural network.

Highly Efficient Convolutional Neural Networks

Resumen de: US20260154533A1

0000 The present disclosure provides directed to new, more efficient neural network architectures. As one example, in some implementations, the neural network architectures of the present disclosure can include a linear bottleneck layer positioned structurally prior to and/or after one or more convolutional layers, such as, for example, one or more depthwise separable convolutional layers. As another example, in some implementations, the neural network architectures of the present disclosure can include one or more inverted residual blocks where the input and output of the inverted residual block are thin bottleneck layers, while an intermediate layer is an expanded representation. For example, the expanded representation can include one or more convolutional layers, such as, for example, one or more depthwise separable convolutional layers. A residual shortcut connection can exist between the thin bottleneck layers that play a role of an input and output of the inverted residual block.

NEURAL NETWORK MODEL FOR SEQUENCE PREDICTION WITH ATTENTION TO ENTITY RELATIONSHIPS

Resumen de: WO2026117281A1

An example formulates a training input for a neural network model with attention to include action data and descriptive content. The action data includes a first entity identifier (ID) and a first sequence of actions associated with the first entity ID. The descriptive content describes a first entity associated with the first entity ID. An action in the first sequence of actions includes an electronic transmission involving the first entity and a second entity. An example uses the training input, including the first entity ID, and a non-standardized tokenizer, to train the neural network model with attention to generate and output a second sequence of actions.

ELECTRONIC DEVICE AND METHOD FOR IDENTIFYING SPEECH SIGNALS FOR INVOCATION IN VARIOUS LANGUAGES, AND NON-TRANSITORY COMPUTER READABLE STORAGE MEDIUM

Resumen de: WO2026116542A1

A processor of an electronic device, according to one embodiment, may acquire: from a first neural network into which a speech signal received through a microphone has been input, a first sequence of portions of the speech signal corresponding to designated frame units; from a second neural network into which designated text has been input, a second sequence of one or more phonetic symbols for the designated text; from a third neural network into which the first sequence and the second sequence have been input, a first dataset indicating the degree to which each of the one or more phonetic symbols corresponds to each of the portions of the speech signal; and, from a pattern discriminator into which the first dataset has been input, predicted phonemes of the portions of the speech signal.

METHOD AND SYSTEM FOR INTEGRATING CONVOLUTIONAL NEURAL NETWORK AND LIGHTWEIGHT ARTIFICIAL INTELLIGENCE FOR ON-DEVICE IMAGE CLASSIFICATION

Resumen de: WO2026116516A1

Provided are a method and system for integrating a CNN with lightweight artificial intelligence for on-device image classification. An on-device integrated artificial intelligence system, according to an embodiment of the present invention, comprises: a first network which is a non-lightweight network that extracts features from an input image; and a second network which is a lightweight network that performs inference on the input image on the basis of the extracted features. Accordingly, the performance of the lightweight artificial intelligence can be complemented by integrating the feature point extraction function of the CNN into a lightweight network in an on-device environment.

INSTRUCTION GENERATION USING ONE OR MORE NEURAL NETWORKS

Resumen de: US20260154955A1

0000 Apparatuses, systems, and techniques are presented for generating instructional text. In at least one embodiment, an instructional video is analyzed to determine logical steps of a process or task demonstrated in that video, and instructive text is generated for those logical steps.

METHOD FOR PREDICTING GROWTH BASED ON GROWTH AGE USING ARTIFICIAL INTELLIGENCE MODEL AND PROVIDING SOLUTION THEREFOR

Nº publicación: US20260154568A1 04/06/2026

Solicitante:

GP CO LTD [KR]
GP CO., LTD.

Resumen de: US20260154568A1

0000 A method for predicting growth on the basis of growth age and providing a solution by using an artificial intelligence model may include the steps of: receiving biometric data of a measurement target; extracting data regarding the predicted age of peak height velocity (APHV), at which the growth velocity is expected to reach the maximum value, by using the biometric data of the measurement target; classifying the growth step of the measurement target into one of multiple growth steps on the basis of the extracted data regarding the predicted APHV; predicting the final height by inputting the extracted data regarding the predicted APHV into a trained neural network; and providing a growth management solution on the basis of the classified growth step and the predicted final height.