2383 results about "Artificial neuronal network" patented technology

A sensor data evaluation device may include an artificial neural network, configured to receive and process first testing data, and output a first testing data output, wherein the first testing data represent sensor data of a first sensor detecting a sensor input; and one or more processors, configured to receive sensor deficiency information representing a deficiency of a second sensor; generate second testing data, wherein the second testing data are based on the sensor deficiency information; implement the artificial neural network to receive and process the second testing data; and output from the output layer a second testing data output; determine a difference between the first testing data output and the second testing data output; and if the difference is outside of one or more predetermined ranges or does not satisfy one or more predetermined criteria, send a signal representing a result of the determined difference.

Methods, systems, and techniques for metadata determination and storage. A large language model that is implemented using at least one artificial neural network receives an initial prompt that includes a query related to the metadata. The metadata is in respect of data that is part of a dataset, and the initial prompt includes context for the query. The large language model determines the metadata in response to the query using the context. Once determined, the metadata is stored in the dataset such that the metadata is associated with the data to which it relates.

An intelligent (self-learning) sensor-aware and / or context-aware subsystem comprising (i) a System-on-a-Chip (SoC), (ii) a radio transceiver, (iii) a microphone, (iv) a voice processing module, (v) a (bio-inspired) neuromorphic event camera or a hyperspectral camera, (vi) a first set of computer implemental instructions in artificial neural networks (ANN) (which may include a transformer model / diffusion model or Poisson flow generative model++ (PFGM++)), (vii) a second set of computer implementable instructions to analyze and interpret contextual data and (viii) an autonomous artificial intelligence (AI) agent is disclosed.

The invention discloses a traffic infrastructure full life cycle carbon emission assessment method based on a knowledge graph. The method comprises the steps that knowledge is acquired through multi-source data acquisition, entities, attributes and relations related to carbon emission are defined, and knowledge graph ontology design and graph construction are carried out; establishing a carbon emission evaluation model based on an artificial neural network on a framework of the knowledge graph through data preprocessing, multi-layer perceptron network structure design and model training; and in combination with the carbon emission evaluation result and the optimization knowledge in the knowledge graph, generating a specific optimization strategy, and performing combinatorial optimization of the optimization strategy to formulate a recommendable optimization scheme for execution of an optimization decision. According to the scheme provided by the invention, carbon emission can be accurately quantified, a scientific and reasonable optimization strategy is generated, dynamic adjustment is supported, and a systematic and intelligent solution is provided for green and low-carbon development of traffic infrastructures and other engineering projects.

Disclosed herein is an artificial neural network-based medical image analysis apparatus for analyzing a medical image based on a medical artificial neural network. The artificial neural network-based medical image analysis apparatus includes a computing system, and the computing system includes at least one processor. The at least one processor is configured to acquire or receive a first analysis result obtained through the inference of a first artificial neural network from a first medical image, to input the first analysis result to a second artificial neural network, to acquire a first evaluation result obtained through the inference of the second artificial neural network from the first analysis result, and to provide the first evaluation result to a user as an evaluation result for the first medical image and the first analysis result.

The present invention discloses a radar scattering field simulation method for a three-dimensional moving target group combined with a neural network, and belongs to the field of electromagnetic simulation technology. The method of the present invention converts the electromagnetic response of a three-dimensional moving target group into the response of the target when it is distributed at different positions, and extracts the three-dimensional numerical Green's function of the randomly distributed target group in the solution domain by changing the spatial distribution of the target group; the three-dimensional numerical Green's function is split into scattering terms and coupling terms, and the artificial neural network is trained separately; when training the artificial neural network, according to the characteristics of the moving target group, its position distribution and scattering terms / coupling terms are input into the artificial neural network as the main features; the scattering terms and coupling terms output by the artificial neural network model are combined into numerical Green's function terms, which can be applied to integral equations to achieve fast and accurate solution of the scattering field of the three-dimensional moving target group.

The invention provides a composite material performance prediction and process optimization method based on a neural network, and the method comprises the steps: firstly collecting multi-source data in the preparation and test process of a composite material, carrying out the preprocessing of the data, screening key feature variables as input variables, constructing a feedforward artificial neural network model, and predicting and outputting the performance indexes of the composite material. And training the model, performing iterative optimization on model parameters, and optimizing composite material process parameters by using the optimized model based on a reverse optimization strategy of a genetic algorithm to obtain an optimal process parameter combination. The invention provides a scientific, efficient and reliable tool for design and optimization of composite materials, and particularly has wide application prospects in high-requirement industries such as aerospace and the like.

A data-centric system for analyzing agricultural crops, consisting of: a data acquisition module configured to capture images of agricultural fields using cameras, unmanned aerial vehicles (UAVs) or sensors, with the sensors collecting data on soil moisture, temperature, light, humidity and pH; a data preprocessing module configured to: resize the acquired images to a standardized dimension suitable for input to a deep learning model; apply noise reduction using a Gaussian filter; improve image contrast through histogram equalization; and perform image magnification through rotation, reflection, and scaling transformations; a feature engineering module configured to extract the following features: color features, which include color histograms, mean, and standard deviation of color channels; texture features using Gray-Level Co-occurrence Matrix (GLCM) properties, which include contrast, dissimilarity, homogeneity, energy, angular moment (ASM), and correlation; shape features, which include contour area, perimeter, aspect ratio, and roundness; and other features, which include the green pixel ratio and edge density; a classification module configured to: implement deep learning-based classification models selected from the group consisting of Support Vector Machine (SVM), Artificial Neural Network (ANN), Convolutional Neural Network (CNN), ResNet18, Random Forest (RF), SegNet, VGGNet, Naive Bayes (NBG), Decision Tree (DT), K-Nearest Neighbors (KNN), and DeepLab; detecting and classifying weed species in the images of agricultural fields; and diagnosing plant diseases based on the features extracted from the images of agricultural fields; an output module comprises a user interface configured to display the classification and recognition results; and a recommendation module configured to suggest treatment solutions for diagnosed plant diseases through the output module's user interface.

Improved placement of memory and functional modules, ‘tiles’, within a tiled processor architecture are disclosed for linear algebra calculations involving vectors and matrices comprising large amounts of data. The improved placement places the data in close proximity to the functional modules performing calculations using the data. These modules enable these calculations to be performed more quickly while using less energy. These modules, in particular, improve the efficiency of the training and application of deep learning and artificial neural network systems. This Abstract and the independent Claims are concise signifiers of embodiments of the claimed inventions. The Abstract does not limit the scope of the claimed inventions.

The invention relates to a factory equipment operation and maintenance decision-making system based on multi-modal data fusion and hybrid reasoning, which comprises a multi-modal data fusion framework and a hybrid reasoning model, and is characterized in that the multi-modal data fusion framework is used for integrating multi-modal data and dynamically weighting fusion features through a cross-modal attention mechanism; the hybrid reasoning model comprises a rule reasoning module, a probabilistic reasoning module and an AI enhancement module, the rule reasoning module generates a generative rule based on an expert knowledge base, the probabilistic reasoning module processes uncertainty data by adopting a dynamic Bayesian modeling equipment state transition probability, and the AI enhancement module introduces an artificial neural network to optimize rule reasoning parameters; and different factory affair scenes are adapted in combination with transfer learning. The method has the advantages that the method is suitable for power stations, power distribution rooms, fans and other scenes, and fault diagnosis, service life prediction and dynamic maintenance strategy generation are combined.

Disclosed is a multi-output regression prediction method for small samples in a laser processing process, which relates to the technical field of machine learning, and uses an adaptive dynamic genetic algorithm (ADGA) and an adaptive moment estimation algorithm (AM) to jointly optimize an artificial neural network (ANN), the algorithms being based on enhanced small samples. Data enhancement is achieved by adding noise to output data in the laser processing process, the ADGA is used to search for optimal parameters of the ANN on a global scale, and the AM is used to optimize the gradients of the optimal parameters, thereby accelerating the convergence speed of the model and enhancing the external prediction stability of the model. When the sample size in laser processing is insufficient, stable and accurate small-sample multi-output prediction can be carried out, thereby improving processing efficiency and reducing production costs in laser processing.

A system includes a neural network architecture. It a new type of neural network able to process statically mapped as well as temporally sequenced information with much better power utilization, data requirements and operational efficiencies. Unlike prior artificial neural network approaches, the present invention includes uniquely defined sets of relationships. The unique use of non-linear input-output mapping functions combined with a time-variant pilot function, and a deterministically bounded, fully-differentiable, nonlinear resonance field subsystem allows the present invention to be readily deployed to work with virtually any neural network architecture / implementation including photonic, opto-acoustic or other variants. This dramatically reduces the size and complexity of virtually any neural network architecture because it offloads what would otherwise need to be done in the form of back / forward propagation trained weights and biases to much simpler, more scalable differentiable input / output mapping functions.

A system for an automated real-time fraud detection based on predictive analytics of call data including a processor of a call analysis server (CAS) node configured to host a machine learning (ML) module and connected to at least one user-entity node over a network and a memory on which are stored machine-readable instructions that when executed by the processor, cause the processor to: capture user call data including a plurality of key elements from the at least one user-entity node; parse the user call data to extract a plurality of classifying features based on the plurality of key elements; query a local calls database to retrieve local historical calls-related data based on the plurality of classifying features; generate at least one feature vector based on the plurality of classifying features and the local historical calls-related data; provide the at least one feature vector to the ML module coupled to an Artificial Neural Network (ANN); receive a plurality of user call ranking parameters from a call predictive model generated by the ML module using outputs of the ANN based on the feature vector; and generate a fraud verdict for the user call based on the plurality of call ranking parameters.

Numerous embodiments are disclosed for splitting an array of non-volatile memory cells in an analog neural memory in a deep learning artificial neural network into multiple parts. Each part of the array interacts with certain circuitry dedicated to that part and with other circuitry that is shared with one or more other parts of the array.

Aspects of this disclosure relate to digital compensators, such as digital predistortion systems. Digital predistortion systems disclosed herein use a neural Volterra approach. Such digital predistortion systems can include a feature processing path comprising a feature artificial neural network, an envelope processing path, multipliers configured to multiply respective output signals of the feature processing path and the envelope processing path, and a combiner configured to generate a combined output signal based on at least output signals of the multipliers. The combined output signal is a digitally predistorted version of an input signal.

The invention discloses a magnetotelluric signal-noise separation method based on an artificial neural network. The magnetotelluric signal-noise separation method comprises the following steps: constructing massive sample libraries conforming to magnetotelluric weak signals and strong interference characteristics; inputting the sample library into a one-dimensional convolutional neural network for training to obtain a corresponding training set and a test set, and defining the network and related training parameters to obtain a training model; performing signal-noise identification on the simulated and actually measured magnetotelluric data by using the obtained mathematical model; noise suppression is only carried out on signals identified as strong interference by using a wavelet soft threshold method; and combining the electromagnetic signal segment after strong interference suppression with the weak electromagnetic signal segment to obtain a reconstructed magnetotelluric signal. According to the method, data segments of strong interference and weak electromagnetic signals can be identified in a more adaptive and high-precision mode, more useful magnetotelluric signals are reserved, and geoelectricity information of a measuring point is better reflected.

A hardware emulator and an emulation system including the hardware emulator are provided. The hardware emulator includes an artificial neural network-based reconstruction model configured to reconstruct dynamics of a dynamical system based on input data and a memristor-based circuit configured to emulate state space representation of the dynamical system based on the reconstruction model.

A method for quantifying lung function using a medical image, comprising the following steps: capturing or receiving a medical image that includes anatomical information for a lung region of a patient; segmenting at least one region of abnormality in the lung region of the medical image using an artificial neural network; and predicting a quantification result with respect to lung function based on the size of the at least one region of abnormality.

The invention discloses a spiking neural network construction method, relates to the technical field of deep learning, and solves the problems that the performance of an existing spiking neural network on a computer vision task is insufficient and the search time is long when an NAS is used. The method comprises the following steps: S1, determining an application scene and a performance requirement of a spiking neural network; s2, collecting data corresponding to the application scene and preprocessing the data; s3, performing training quantization on the artificial neural network, and then converting the trained artificial neural network into a spiking neural network according to an equivalent mapping relation between spatial quantization of the artificial neural network and time quantization of the spiking neural network; s4, searching a pulse neural network with the best performance according to a peak neural structure search algorithm; s5, deploying the spiking neural network with the optimal performance to a specified application scene; according to the method, the optimal SNN architecture is automatically searched by introducing a spiking neural structure search method, so that the requirement on manual design is reduced, and the design efficiency and the network performance are improved.

A computer-implemented method produces latent representations of text prompts created for use with a text-to-image diffusion model. Training images are generated by providing the latent representations to a first artificial neural network implementing a denoising process of the text-to-image diffusion model. A machine-learned modality inversion module is trained. The training includes performing training iterations for training data pairs, each training data pair being comprised of one of the training images and one of the text prompts. Each training iteration for each training data pair includes: providing the one of the training images of the training data pair to a pre-trained classifier configured to generate alternate conditioning information based upon the one of the training images, converting, by the machine-learned modality inversion module, the alternate conditioning information into an alternate latent representation, and updating parameters of the machine-learned modality inversion module based upon differences between the alternate latent representation and one of the latent representations of the one of the text prompts.

The invention provides a tidal current energy water turbine intelligent design method based on a generator and a predictor and a water turbine structure optimization design framework. According to the method, a compact generator is obtained by training a variational self-encoding generative adversarial network and is used for parametric modeling of tidal current energy water turbine rotor blades, and an efficient performance predictor based on the combination of an artificial neural network and transfer learning is established to realize intelligent prediction of blade performance. The generator and the predictor are coupled with the multi-target genetic algorithm, efficient design of the tidal current energy water turbine blade is achieved, the performance of the tidal current energy water turbine blade can be rapidly predicted, and then efficient optimization design of the tidal current energy water turbine is achieved.

The invention relates to an underwater target detection method based on a pulse neural network, which can realize underwater target detection with high precision and low power consumption. By fusing the cross-stage partial network and the YOLO architecture, the problem of pulse degradation is effectively solved, and the feature extraction capability of the model is enhanced; in order to solve the problem of underwater noise interference, the pulse-based underwater image denoising method is designed, only integer addition is used in the method, a pulse neural network structure can be seamlessly embedded, and the quality of a feature map is enhanced; in order to solve the problem that a traditional normalization method is low in precision in the spiking neural network, separate batch normalization is provided, by independently normalizing a feature map in multiple time steps and optimizing a residual structure, the time dynamic state of the SNN can be effectively captured, and the detection precision is improved. The network shows excellent performance in underwater target detection, and compared with an artificial neural network of the same scale, the network has higher performance and lower energy consumption.

An inspection method for a lithium secondary battery can improve the reliability, accuracy, and reproducibility of inspection results by advancing a learning of an artificial neural network according to one or more divided regions of one or more surfaces of the lithium secondary battery and one or more types of defects occurring in each region.

A set of values can be selected from a plurality of fields of a table in a database. At least one adverb or adjective can be selected for the set of values. Join paths for values in the set of values can be determined. A structured query language pattern can be determined based, at least in part, on at least one value in the set of values, the at least one adverb or adjective for the set of values, and the join paths for the set of values. The structured query language pattern can be stored to first training data configured, at least in part, for use in machine learning to train a text-to-SQL model, the text-to-SQL model comprising a first artificial neural network and configured to convert first natural language text to a first structured query language query.

The invention relates to the technical field of intelligent lighting control systems, in particular to an underground garage lighting control system and method based on artificial neural network PID and a storage medium. Comprising a brightness sensor, a traffic flow sensor, an illuminance sensor, a temperature and humidity sensor and a microwave speed measurement sensor which are arranged inside and outside a garage; the intelligent network connection lamp comprises an intelligent gateway unit, an LED dimming driving unit and an LED lamp, and the intelligent gateway unit is used for receiving the control instruction and communicating with the upper computer; and the upper computer is used for receiving the data of the detection unit, generating an optimal illumination control curve based on a garage comfort illumination model and a rendezvous zone fuzzy control model, and adjusting the illumination of lamps in each zone in real time through an RBF neural network PID (Proportion Integration Differentiation) algorithm. According to the method, self-learning optimization is achieved, PID parameters are adjusted online through the neural network RBF, the method adapts to long-term dynamic influences such as differences of different types of garages, seasonal changes and equipment aging, and optimal control is given.

The invention provides an ion beam polishing multi-working-condition removal function obtaining method which comprises the steps that according to a to-be-machined optical surface shape and a planned movement track of an ion source relative surface shape, the change range of an ion beam incident angle and the change range of a machining distance in a polishing machining task are determined; carrying out limited times of single-point etching experiments, and detecting the surface shape of the workpiece after the single-point etching experiments; calculating a processing amount distribution, and calculating a removal function considering the incident angle of the ion beam and the processing distance based on the processing amount distribution data and the processing time; training an artificial neural network by taking the removal function as training data; and on the basis of the trained neural network, obtaining a removal function fitting parameter pair according to the actual processing task requirement, and performing inverse processing on the removal function fitting parameter pair to obtain a removal function of the actual processing point. According to the removal function obtained through the method, the influence under multiple working conditions is considered, and the current high-precision machining requirement for the high surface slope or various non-planar optical surface shapes can be better met.

A computer-implemented method generates, using training frames of training image data in combination with a first set of data derived from the training frames of training image data, a set of fine-tuning weights for a pre-trained diffusion model implemented by a first artificial neural network. A first digital signature is generated based upon values of the fine-tuning weights. The values of fine-tuning weights and the first digital signature are sent to a computing device configured to, after verifying the first digital signature, use the fine-tuning weights to establish a specialized diffusion model implemented by a second artificial neural network.

A method training an artificial neural network (ANN) on a remote host computes, using a trusted process deployed in a trusted execution environment (TEE) on the remote host, a key-pair for a homomorphic encryption scheme and shares, by the trusted process, the public key (PK) of the key-pair with an untrusted process deployed on the remote host. The method splits the training procedure of the ANN between the untrusted process and the trusted process, wherein the untrusted process computes encrypted inputs to neurons of the ANN by means of the homomorphic encryption scheme, while the trusted process computes outputs of the neurons based on the respective encrypted inputs to the neurons as provided by the untrusted process.

The invention relates to a silicon photonic interlayer coupling structure design method based on an artificial neural network, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring a plurality of interlayer coupling structure data sets, performing normalization processing on the interlayer coupling structure data sets, and dividing the data sets; constructing a back propagation neural network; in the design process, a self-adaptive genetic algorithm is used for optimizing the weight and the threshold value of the back-propagation neural network; using a training set to train the back propagation neural network model; coupling structure parameters are input into the trained model, a coupling efficiency value is output, and whether the coupling structure meets design requirements or not is judged according to the coupling efficiency value; and continuously adjusting the structural parameters until the design requirements are met. According to the method, the design time of the interlayer coupling structure is shortened, the design difficulty of the interlayer coupling structure is reduced, and automation and intelligence of the design of the interlayer coupling structure are realized.

Systems, devices, and methods related to a Deep Learning Accelerator and memory are described. For example, an integrated circuit device may be configured to execute instructions with matrix operands and configured with random access memory. The random access memory is configured to store input data from a sensor, parameters of a first portion of an Artificial Neural Network (ANN), instructions executable by the Deep Learning Accelerator to perform matrix computation of the first portion of the ANN, and data generated outside of the device according to a second portion of the ANN. The Deep Learning Accelerator may execute the instructions to generate, independent of the data from the second portion of the ANN, a first output based on the input data from the sensor and generate a second output based on a combination of the data from the sensor and the data from the second portion of the ANN.