Neural networks

STAGED NEURAL NETWORK CLASSIFICATION

Absstract of: US20260178882A1

Events may be classified in a staged manner using one or more neural networks and explicit classifiers. For example, a method may be performed which comprises classifying an event using a first explicit classifier, a neural network classifier, and a second explicit classifier. In such a method, classifying the event using the first explicit classifier may provide an initial classification for the event, and that initial classification may be a basis for classifying the event using the neural network classifier. Similarly, classifying the event using the neural network classifier may provide a neural network classification for the event, and the event may be classified using the second explicit classifier on the basis of the neural network classification. When the event is processed by the second explicit classifier, this may provide an output classification, and that output classification may be used as a basis for processing the event.

METHOD AND APPARATUS FOR IMPROVING ACCURACY OF INTEREST POINT DETECTION UNDER DIFFERENT LIGHT CONDITIONS

Absstract of: US20260178905A1

According to at least one embodiment, a computer-implemented method of training a neural network for mapping an indoor environment includes: training the neural network in a first stage using a first dataset based on synthetic shapes; and training the neural network in a second stage using a second dataset based on real photographs. The method further includes, for each object of a plurality of objects, collecting a plurality of images of the object, wherein the plurality of images of the object are respectively produced under different lighting conditions; and training the neural network in a third stage using the plurality of images of the object.

SYSTEM AND METHOD FOR ASSISTING WITH THE NAVIGATION OF A MOBILE SYSTEM

Absstract of: US20260178899A1

A method for assisting navigation of a mobile system includes obtaining an optical image of a scene acquired by a camera, obtaining a 3D point cloud of the scene acquired by a range-finder, projecting, in 2D into a reference frame of the camera, the 3D points of the 3D point cloud and a measurement uncertainty to provide a depth image and an uncertainty mask of the depth image, determining a semantic map, a depth map and a confidence map of the depth map from the optical image by processing the optical image, the depth image and the uncertainty mask a first convolutional neural network which includes a succession of convolutional layers, each including first to third convolution blocks that estimate respectively a semantic attribute map, a depth attribute map and a confidence attribute map, and determining a scene traversability map by merging the semantic, depth, and confidence maps.

METHOD AND DEVICE FOR PERFORMING FEDERATED LEARNING IN SATELLITE COMMUNICATION SYSTEM

Absstract of: US20260178932A1

0000 A method for performing federated learning in a satellite communication system includes receiving, at a base station device, association information from each of all satellites within a coverage, setting, one or more satellites available for training among the connected satellites as training satellites based on the received association information, generating, training information for each of the training satellites based on the received association information, transmitting, the generated training information and a global neural network model to each of the training satellites, receiving, each of local neural network models whose training has been completed from the training satellites, and transmitting, the received local neural network models to a server.

SYSTEM AND METHOD FOR AUTOMATED MULTI-SPEAKER AND MULTI-LINGUAL SPEECH ANALYSIS

Absstract of: US20260178838A1

Exemplary system and methods use a combination of application modules and neural network architecture for multi-speaker and multi-language speech analysis. The exemplary system can receive a natural language input, which it decomposes into plural segments. A sub-group of the plural segments are accumulated in a buffer where each segment representing a period during which voice activity is detected. The sub-groups are analyzed for voice activity of multiple speakers and one or more text segments are generated based on the speakers. A semantic vector for each text segment is generated and stored in vector memory. Relevant data associated with each semantic vector is retrieved from the vector memory based on a similarity measure; and a response including specified information extracted from the one or more text segments is generated based on at least the relevant data.

ANTI-OVERFITTING LIGHTWEIGHT ANOMALY DETECTION NEURAL NETWORK MODEL RETRAINING METHOD

Absstract of: US20260178961A1

Disclosed in the present invention is a lightweight anomaly detection neural network model retraining method with anti-overfitting, which retrains an anomaly detection model based on depth variational autoencoders. When a data distribution changes, a conditional distribution of a hidden state and reconstructed data samples obtained by an encoder and a decoder of the depth variational autoencoders will also change. The present invention uses a mapping function to adjust the conditional distribution of the hidden state and the reconstructed data obtained by the calculation of an old model to adapt to a new data distribution. The mapping function has simple and convex characteristics, and can ensure a fast convergence rate and light overhead in a retraining process on a premise of using a loss function form defined by the present invention. In addition, the present invention provides a rumination module for data enhancement of new observation data to solve a problem of insufficient new observation sample data in an initial period when cloud service characteristics change.

NEURAL NETWORK CIRCUIT AND NEURAL NETWORK OPERATION METHOD

Absstract of: US20260178895A1

A neural network circuit having multiple neural network operation cores having convolution operation circuits that perform convolution operations and quantization operation circuits that perform quantization operations, wherein the multiple neural network operation cores are connected so as to be able to input and output data.

IMAGE GENERATION USING ONE OR MORE NEURAL NETWORKS

Absstract of: US20260179197A1

Apparatuses, systems, and techniques are presented to generate one or more images. In at least one embodiment, one or more neural networks are used to generate one or more images based, at least in part, on one or noise values.

SYSTEMS AND METHODS FOR AUTOFOCUS AND AUTOMATED CELL COUNT USING ARTIFICIAL INTELLIGENCE

Absstract of: US20260177470A1

Systems and methods for autofocus using artificial intelligence include (i) capturing a plurality of monochrome images over a nominal focus range, (ii) identifying one or more connected components within each monochrome image, (iii) sorting the identified connected components based on a number of pixels associated with each connected component, (iv) generating a focus quality estimate of at least a portion of the sorted connected components using a machine learning module, and (iv) calculating a target focus position based on the focus quality estimate of the evaluated connected components. The calculated target focus position can be used to perform cell counting using artificial intelligence, such as by (i) generating a seed likelihood image and a whole cell likelihood image based on output—a convolutional neural network and (ii) generating a mask indicative quantity and/or pixel locations of objects based on the seed likelihood image.

IMAGE ENHANCEMENT USING ONE OR MORE NEURAL NETWORKS

Absstract of: US20260179180A1

Apparatuses, systems, and techniques are presented to generate images with one or more visual effects applied. In at least one embodiment, one or more visual effects are applied to one or more images having a resolution that is less than a first resolution and those visual effects approximated for one or more images having a resolution that is greater than or equal to the first resolution.

SYSTEMS AND METHODS FOR DIAGNOSING AND/OR TREATING PATIENTS

Absstract of: US20260179222A1

Devices, systems, and methods are provided for recognizing, diagnosing, mapping, sensing, monitoring and/or treating selected areas within a patient's body. The systems, devices and methods may be used to map, detect and/or quantify images and/or physiological parameters collected from the patient. One such system comprises an optical imaging device, such as an endoscope, and a processor coupled to the imaging device. The processor includes a software application configured to recognize the images captured by the optical imaging device and determine if the tissue contains a medical condition and may include an artificial neural network configured to develop at least one set of computer-executable rules useable to recognize the medical condition in the captured tissue images. The systems, devices and methods provided herein allow for a more objective and comprehensive inspection of the targeted areas within a patient so as to improve the diagnosis and ultimate treatment of patients.

SCALABLE FORM MATCHING

Absstract of: US20260178825A1

0000 Disclosed are a method and apparatus for determining a given template of a form used by a filled in instance of that type of form from amongst a great number of form templates (a hundred or more). The given instance is evaluated by a neural network that has been trained by a single example of each template in order to reduce the total number of templates down to a manageable amount. Given a list of closest matching templates, the instance is aligned to each of the closest matching templates. The comparison generates a match score. The form template having the greatest match score is the correct form template. Filtering the instance through a one-shot learning neural network before performing a precise comparison enables the process to scale to any number of template forms.

MODEL STRUCTURE, METHOD FOR TRAINING MODEL, IMAGE ENHANCEMENT METHOD, AND DEVICE

Absstract of: US20260179360A1

Embodiments of this application disclose a model structure, a method for training a model, an image enhancement method, and a device, and may be applied to the computer vision field in the artificial intelligence field. The model structure includes: a selection module, a plurality of first neural network layers, a segmentation module, a transformer module, a recombination module, and a plurality of second neural network layers. The model overcomes a limitation that the transformer module can only be used to process a natural language task, and may be applied to a low-level vision task. The model includes the plurality of first/second neural network layers, and different first/second neural network layers correspond to different image enhancement tasks. Therefore, after being trained, the model can be used to process different image enhancement tasks.

CREDIT ELIGIBILITY PREDICTOR

Absstract of: US20260179153A1

Aspects extract, from payroll data of employees of an organization, data historically associated to previous instances of certified tax credit eligibility; normalize the extracted data with respect to data type and data value; generate from the normalized extracted data via a neural network classifier multi-class outputs for each employee that indicate strengths of likelihood that each employee is currently eligible for each of a plurality of different tax credits; filter the normalized extracted data by removing portions associated to employees indicated within the multi-class outputs as having no currently eligible likelihood for the different tax credits, thereby generating a remainder set of normalized extracted data associated to remainder eligible ones of the employees; and prioritize application for the tax credits for the remainder eligible employees as a function of respective values and likelihoods of eligibility within the remainder set of normalized extracted data.

GENERATING AUTOMATED ASSISTANT RESPONSES AND/OR ACTIONS DIRECTLY FROM DIALOG HISTORY AND RESOURCES

Absstract of: US20260179618A1

Training and/or utilizing a single neural network model to generate, at each of a plurality of assistant turns of a dialog session between a user and an automated assistant, a corresponding automated assistant natural language response and/or a corresponding automated assistant action. For example, at a given assistant turn of a dialog session, both a corresponding natural language response and a corresponding action can be generated jointly and based directly on output generated using the single neural network model. The corresponding response and/or corresponding action can be generated based on processing, using the neural network model, dialog history and a plurality of discrete resources. For example, the neural network model can be used to generate a response and/or action on a token-by-token basis.

MACHINE-LEARNING-ASSISTED SELF-IMPROVING OBJECT-IDENTIFICATION SYSTEM AND METHOD

Absstract of: US20260179351A1

A system and method of identifying and tracking objects comprises registering an identity of a person who visits an area designated for holding objects, capturing an image of the area designated for holding objects, submitting a version of the image to a deep neural network trained to detect and recognize objects in images like those objects held in the designated area, detecting an object in the version of the image, associating the registered identity of the person with the detected object, retraining the deep neural network using the version of the image if the deep neural network is unable to recognize the detected object, and tracking a location of the detected object while the detected object is in the area designated for holding objects.

PROCESSING METHOD FOR NEURAL NETWORK MODEL, AND SECURE ELEMENT AND COMPUTING APPARATUS

Absstract of: WO2026129814A1

Provided in the embodiments of the present application are a processing method for a neural network model, and a secure element and a computing apparatus. The computing apparatus comprises: a storage element, which is used for storing a first neural network code for the first model inference of a neural network model and at least some first network parameters for the first model inference; a secure element, which is used for storing the remaining first network parameters for the first model inference, and/or storing a second neural network code and second network parameters for the second model inference of the neural network model, and executing the second model inference on the basis of the second network parameters and the second neural network code; and a general-purpose computing element, which is used for executing the first model inference on the basis of the first neural network code and the first network parameters. The embodiments of the present application can prevent a neural network model from being physically attacked, thereby improving the security of the neural network model.

LARGE LANGUAGE MODEL INFERENCE APPARATUS BASED ON COMPUTE-IN-MEMORY, INFERENCE SYSTEM, AND ELECTRONIC DEVICE

Absstract of: WO2026129865A1

The present application relates to the field of artificial intelligence, and discloses a large language model (LLM) inference apparatus based on compute-in-memory, an inference system, and an electronic device. The inference apparatus comprises: a storage layer at least used for storage; and a computation layer at least used for computation. The computation layer and the storage layer are stacked by means of hybrid bonding. The computation layer comprises a neural network accelerator based on compute-in-memory. The neural network accelerator comprises an in-memory computing matrix. The in-memory computing matrix is used for performing neural network computation on input feature data and weights from the storage layer. The computation layer is further used for being electrically connected to a main control chip that controls the inference apparatus. The computation layer is further used for performing prefill processing of LLM inference and transmitting data, which is obtained after the prefill processing, to the main control chip for decoding processing of LLM inference, such that the prefill processing is separated from the decoding processing. The inference apparatus provided by the present application supports high bandwidth, has high computational power and low power consumption, and can also resolve the heat dissipation problem of existing LLM inference apparatuses.

OBJECT CLASSIFICATION

Absstract of: US20260179374A1

0000 In various examples, multilabel hierarchical classification of objects for autonomous systems and applications is described herein. Systems and methods are disclosed that use one or more neural networks to classify objects, such as traffic signs, using multilabel classification and/or hierarchical classification. For instance, a multilabel subnetwork of the neural network(s) may classify an object based at least on one or more attributes associated with the object. As such, the output from the multilabel subnetwork may include at least a classification associated with the object and an attribute classification(s) associated with the object. A hierarchical subnetwork of the neural network(s) may also classify the object using one or more class labels, where a class label indicates another classification and/or a class group associated with the object. The systems and methods may then use the classification, the attribute classification(s), and/or the class label(s) to determine a final classification associated with the object.

APPARATUS AND METHOD FOR PROCESSING SENSOR DATA, SENSOR SYSTEM

Absstract of: US20260179315A1

0000 Processing of sensor data from sensor(s). The sensor data are provided as an unordered point cloud. The points of the unordered sequence are then converted into a regular structure using a point-processing neural network and made available for further processing. A transfer device is configured to receive a group of input data elements from the sensors. Each input data element of this group of input data elements includes a point that specifies at least one position. The transfer device also includes a point-processing neural network. This point-processing neural network is configured to map the points of the group of input data elements to a regular output data structure. A processing device is configured to detect an object and/or ascertain properties of an object using the regular output data structure. For the conversion of points of an unordered point cloud to a regular structure, a point-processing neural network is provided.

METHOD FOR DETERMINING SENSOR POSE BASED ON VISUAL DATA AND NON-VISUAL DATA

Absstract of: US20260179249A1

The present disclosure relates to a method for determining position and orientation of a visual sensor and within an environment. The method comprises acquiring a training set of visual data of the environment and an object arranged therein, training an interpolation neural network for estimating one or more synthetic poses using the first set of visual data, and training a convolutional neural network with the first set of visual data. The method comprises acquiring an inspection set of visual data of the environment and an object arranged therein, estimating a coarse pose with the convolutional neural network, and predicting a synthetic image associated with the coarse pose with the interpolation neural network. The method may be performed with data obtained from non-visual sensors.

GENERATING AUDIO USING AUTO-REGRESSIVE GENERATIVE NEURAL NETWORKS

Absstract of: EP4765100A2

0001 Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a prediction of an audio signal. One of the methods includes receiving a request to generate an audio signal; obtaining a semantic representation of the audio signal; generating, using one or more generative neural networks and conditioned on at least the semantic representation, an acoustic representation of the audio signal; and processing at least the acoustic representation using a decoder neural network to generate the prediction of the audio signal.

DATA-FREE POST-TRAINING QUANTIZATION METHOD AND APPARATUS, DEVICE, AND STORAGE MEDIUM

Absstract of: EP4764966A1

This application provides a data-free post-training quantization method and apparatus, a device, and a storage medium, and relates to the field of neural network technologies. The method includes: obtaining data distribution input by a user, where the data distribution is distribution to which an input activation value of each network layer of a floating-point model conforms; inputting random data into the floating-point model to obtain the input activation value of each network layer; performing statistical analysis on the input activation value of each network layer based on the data distribution, to obtain a data range of the input activation value of each network layer; determining a quantization parameter of the input activation value of each network layer based on endpoint values of the data range; and during inference by using the floating-point model, performing, by using the quantization parameter of the input activation value of each network layer, quantization processing on the input activation value generated during inference of each network layer. According to the solution of this application, quantization processing can be performed on an input activation value in a data-free manner.

DATA PROCESSING METHOD AND APPARATUS

Absstract of: EP4765041A1

A data processing method is provided. The method is applied to image processing and includes: obtaining first data collected by an image sensor; and obtaining spectral information based on the first data by using a neural network model, where the neural network model includes an attention module, and the attention module is configured to determine an attention matrix based on input data, and perform an attention operation based on the attention matrix, where the attention matrix is obtained by performing a first fusion operation on correlation information between different channels of the input data and correlation information of the channels. In this application, a degree of correlation between the different channels and a degree of correlation of the channels may be fused, so that the attention matrix can model both correlation and particularity between the different channels, thereby improving accuracy of spectral signal reconstruction.

GENERATIVE NEURAL NETWORKS WITH EFFECTIVE AUDIO TOKEN PROCESSING

Nº publicación: WO2026128699A1 18/06/2026

Applicant:

GDM HOLDING LLC [US]
GDM HOLDING LLC

Absstract of: WO2026128699A1

Methods, systems, and apparatuses, including computer programs encoded on computer storage media, for processing a received network input that includes audio data using a generative neural network to generate an output sequence that represents a transcription of speech included in the audio data. Then processing the output sequence of output tokens to generate a speech recognition output. One of the described techniques include training the generative neural network to generate outputs that interleave audio and text tokens. Another of the described techniques includes receiving and generating audio at the same time step.