308 results about "Neural network system" patented technology

A computer-implemented method of training a neural network system comprising a visual encoder neural network and a text encoder neural network is provided The method comprises obtaining a plurality of training data items (each training data item comprising an image and associated text defining a sequence of text tokens) and at each of a plurality of training steps processing at least one of the training data items by: processing pixels of the image in the training data item using the visual encoder neural network to generate a set of patch embeddings for the image; processing the sequence of text tokens using the text encoder neural network to generate a sequence of token embeddings, processing the set of patch embeddings and the sequence of token embeddings to generate a set of language-aware patch embeddings (based on similarities between patch embeddings and token embeddings), and training at least the visual encoder neural network by backpropagating gradients of a contrastive objective function evaluated over the language-aware patch embeddings and the sequence of token embeddings.

A neural network system is provided. The system includes an autoencoder configured to encode input data into a latent space representation; a generator neural network configured to receive a noise vector and the latent space representation and output a set of routing coefficients; a discriminator neural network configured to evaluate the effectiveness of the routing coefficients by measuring the performance of a capsule network utilizing said routing coefficients; and a capsule network comprising a first capsule layer and a second capsule layer, wherein the routing coefficients are used to dynamically route outputs from the first capsule layer to the second capsule layer.

A neural network system is trained to generate textual reports (that is, medical reports, such as radiology reports) from one or more medical images, by fine-tuning a pre-trained neural network system (a visual language model, "VLM") operative, upon receiving an input comprising at least one image and a textual input, to generate a value indicative of a predicted likelihood of one or more candidate text continuations of the textual input. The fine-tuning of the neural network system is performed to reduce the value of a cost function which includes a first prediction cost term based on a first training database including first training datasets of at least one medical image and an associated text report, the first training datasets corresponding to first individuals. The first prediction cost term further includes a cost value for each individual, inversely dependent on a likelihood value of the associated textual report, conditioned on the at least one medical image, and created by the neural network system.

A neural network system that is configured to learn a representation of data item, such as an image, audio, or text data item, through a self-supervised learning process. Implementations of the system couple two learning processes, an object discovery learning process and an object feature representation learning process. In implementations the object discovery learning process assists the object feature representation learning process in self-supervised learning of object feature representations, and the object feature representation learning process is used to improve the object discovery learning process.

A method is provided for processing data in a neural network system. The method includes receiving input data; processing the input data through a first set of neural network layers configured to perform data processing using MatMul-free techniques to produce intermediate data; further processing the intermediate data through a second set of neural network layers configured to simulate spiking neural network (SNN) functionalities using MatMul-free techniques; and outputting a result based on the processed data from the second set of neural network layers.

Systems and methods for performing image modification. In particular, the system can, using a rectified flow neural network, perform an image inversion and image editing process to generate a modified image that has been modified according to a conditioning input received by the system.

Systems and methods are provided for analog hardware realization of neural networks. The method includes obtaining a neural network topology and weights of a trained neural network. The method also includes transforming the neural network topology to an equivalent analog network of analog components including operational amplifiers and resistors. Each operational amplifier represents an analog neuron of the equivalent analog network, and each resistor represents a connection between two analog neurons. The method also includes computing a weight matrix based on the weights of the trained neural network. The method also includes generating a resistance matrix for the weight matrix. The method also includes pruning the equivalent analog network to reduce the number of operational amplifiers or the resistors, based on the resistance matrix, to obtain an optimized analog network of analog components.

The invention relates to the technical field of geological early warning, in particular to a dangerous rock mass instability analysis method, system and equipment based on a space-time diagram neural network, by fusing unmanned aerial vehicle LiDAR, multispectral data, meteorological radar data and the space-time diagram neural network (ST-GNN), the system realizes sub-meter spatial resolution and minute-level time response, and the stability of dangerous rock mass instability analysis is improved. The four-dimensional (time and space) analysis result of the instability probability of the dangerous rock mass is obtained through high-precision space-time modeling, the problems that a traditional geological disaster early warning system is low in resolution ratio, slow in response and high in misinformation are solved, the comprehensiveness, accuracy and reliability of instability prediction of the dangerous rock mass are improved, and the early warning effect is good. And full-chain intelligent closed-loop management of real-time data acquisition-dynamic prediction-early warning push-feedback optimization is supported, the emergency decision time is shortened by real-time rainfall superposition risk thermodynamic diagrams, and the attenuation rate of long-term prediction precision is reduced by dynamically fusing newly added geological data and instability events through incremental learning.

Systems and techniques are described herein for training and using a machine-learning model (e.g., a neural network). For example, a computing device can: process, using a first trained neural network, data specific to a user to obtain intermediate activation data representing the data, the first trained neural network comprising a plurality of neural network layers; process, using a second trained neural network, the intermediate activation data to generate an output representing the data, the second trained neural network comprising a subset of neural network layers from the plurality of neural network layers of the first trained neural network; determine a loss based on the output; and update parameters of the second trained neural network based on the loss.

The invention belongs to the field of hardware neural network design, and particularly discloses a CMOS-based attractor recurrent neural network system and an implementation method thereof.A neuron calculation module receives an external input high-level signal and generates an initial synaptic high-level signal; integration and activation calculation are carried out based on the current output by the synaptic calculation module, and whether a synaptic high-level signal is generated or not is determined according to an activation state; the synaptic storage module writes the non-zero weight into a synaptic weight SRAM (Static Random Access Memory) array; the synaptic calculation module performs product operation on the synaptic high-level signal and the synaptic weight in the synaptic weight SRAM array, and outputs current; when the STDP updating module is in a neuron training mode, the STDP updating module updates the synaptic weight SRAM array according to the synaptic high-level signal; and the group state module outputs a neuron sequence number corresponding to the maximum activation frequency. A cyclic connection structure of the attractor neural network is formed, and the real-time cognitive behavior requirement of the brain is met.

Systems and methods for using a distributed computing system to train a large neural network to perform a machine learning task. A shared set of trainable parameters is maintained in a shared data store, and each of a geographically distributed set of workers updates their trainable parameters using a shard training dataset. There are two optimization processes: an outer optimization process, and an inner optimization loop that is executed by each worker independently and in parallel tens, hundreds, or thousands of times. The workers can have different computing capabilities and can be geographically distant from one another, and the communications bandwidth used by the system can be two or three orders of magnitude less than that of other systems.

A computer-implemented method for training an anomaly detection neural network system comprising an encoder and a decoder is described. The method includes receiving training data comprising a plurality of training examples, each training example including a training audio waveform and a machine identity (ID); processing the training audio waveform to extract training audio features; receiving an environmental noise audio waveform; processing the environmental noise audio waveform to extract noise features; generating augmented features by combining the extracted training audio features and the noise features; processing, using the encoder, the augmented features to generate latent embeddings; processing, using the decoder, the latent embeddings to generate reconstructed audio features; processing, using a convolutional neural network, the augmented features to generate a predicted machine ID probability distribution; and adjusting, through backpropagation, the current values of the parameters of the encoder and the decoder to minimize an objective function.

A computer-implemented method for generating and ranking source code for performing a task is described. The method includes: receiving input data comprising a task description, a code generation prompt and a test case generation prompt; processing the input data using at least one trained code generation neural network to generate a plurality of code solutions and a plurality of test cases; for each code solution, executing the set of candidate source code on the test inputs of the plurality of test cases to generate a plurality of execution outputs; clustering the plurality of code solutions into a plurality of clusters; computing an interaction matrix that specifies functional overlap between the plurality of clusters; determining, for each cluster, a score based the interaction matrix; and ranking the plurality of clusters based on the scores of the plurality of clusters.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing inputs using a neural network system that includes one or more pre-normalized layers or one or more regularization normalization layers.

Systems and methods, implemented as computer programs on one or more computers in one or more locations, for generating a sequence of data elements using a neural network comprising a sequence of attention neural network layers. The sequence comprises a respective continuous valued data element at each position in a sequence of positions. Implementations of the described techniques remove the need for discrete tokens and fixed, finite vocabularies.

The present invention discloses a memristor pulse neural network system and method, which belongs to the field of artificial intelligence technology. The memristor pulse neural network system provided by the present invention uses a bionic design concept to accurately simulate the synaptic structure of a biological neural network by connecting pairs of memristors of the same level in series; based on the positive and negative Hebbian plasticity rules, an adaptive learning mechanism is implemented to strengthen the computing function of the memristor's storage and computing performance; a cascaded scalable control architecture combining master control and slave control is adopted to give the system a high degree of scalability. According to the needs of specific tasks, users can flexibly adjust the scale of the neural network to achieve unlimited cascade expansion, thereby easily coping with various complex scenarios. In addition, the built-in ADC acquisition module provides a real-time feedback mechanism, which can improve the accuracy of system monitoring, speed up the response speed, and provide valuable data support for system optimization and adjustment.

Systems and techniques are provided for multi-stage training of a multi-network system. An example method can include training, using training data, a first neural network during a first training stage; generating, by the first neural network, one or more outputs; based on a determination that the first training stage and training of the first neural network are complete, providing the one or more outputs to a second neural network that has an input data dependency comprising data generated by the first neural network; and based on the determination that the first training stage and training of the first neural network are complete, training, using the one or more outputs from the first neural network, the second neural network during a second training stage initiated after completion of the first training stage.

Systems and method are provided that are directed to tuning a hyperparameter associated with a small neural network model and transferring the hyperparameter to a large neural network model. At least one neural network model may be received along with a request for one or more tuned hyperparameters. Prior to scaling the large neural network, the large neural network is parameterized in accordance with a parameterizing scheme. The large neural network is then scaled and reduced in size such that a hyperparameter tuning process may be performed. A tuned hyperparameter may then be provided to a requestor such that the hyperparameter can be directly input into the large neural network. By tuning a hyper parameter using a small neural network, significant computation cycles and energy may be saved.

The application discloses a neural network system, a method, and a device for processing floating-point number. A self-defined floating-point number is obtained, wherein the self-defined floating-point number comprises a sign field, an exponent field, and a mantissa field, and a value of the self-defined floating-point number is determined by bit of the sign field, bits of the exponent field, bits of the mantissa field, and a bias value, wherein the bias value is determined by a total bit number of the exponent field. The self-defined floating-point number is applied to numerical calculations.

The present disclosure relates to a method and system for performing packet loss concealment using a neural network system. The method comprises obtaining a representation of an incomplete audio signal, inputting the representation of the incomplete audio signal to an encoder neural network and outputting a latent representation of a predicted complete audio signal. The latent representation is input to a decoder neural network which outputs a representation of a predicted complete audio signal comprising a reconstruction of the original portion of the complete audio signal, wherein said encoder neural network and said decoder neural network have been trained with an adversarial neural network.

A RELU neuron chip circuit belongs to the field of chip circuits, and is characterized by comprising a vector summation circuit, a shift circuit, a subtraction circuit, a logic judgment circuit, an input port and an output port, the input ports comprise a data input port, a weight input port, a bias data input port and a data counting port; the output port comprises a data output port and a logic output port; by directly realizing the function of the neurons on the chip circuit, when a neural network system calls a certain neuron to carry out corresponding function calculation, data input and output can be directly carried out at the bottommost circuit level, so that a large amount of data cross-layer conversion time is saved. Only after the whole neural network completes one time of complete training or judgment operation, the operation result of the bottom layer circuit can be transmitted to the high layer from the bottom layer circuit in a cross-layer mode and displayed in a human eye recognizable mode. Therefore, the overall operation performance of the neural network system is improved.

The invention provides a neural network system, a reasoning method based on the neural network system and a storage medium, and the method comprises the steps: inputting an input matrix corresponding to reasoning input information into a hidden layer, sequentially transmitting the input matrix backwards through N network layers, and when a difference value between an input conversion vector and an output conversion vector of the kth network layer is smaller than a first threshold value, outputting the input matrix; the output matrix of the kth network layer jumps to the last network layer, otherwise, the output matrix is sequentially transmitted to the next network layer; and when the difference value between the input conversion vector and the output conversion vector of the last network layer is greater than a second threshold value, triggering the output matrix of the last network layer to be forwards transmitted to the front-row network layer, continuing to perform backward reasoning, outputting the output matrix of the last network layer as a reasoning matrix, and obtaining reasoning result information according to the reasoning matrix. Thus, based on feedforward acceleration and feedback thinking of the neural network, the input reasoning information can be processed more flexibly, and the efficiency and accuracy of reasoning (prediction) are improved.

For neural network acceleration, a datapath can be configured to implement a convolution computation. A control unit can be configured to coordinate operations of the datapath to implement the convolution computation based on coded instructions representative of a neural network system. The control unit can be configured to command the datapath to convolve at least one input feature element of a set of input feature elements of at least one input feature map with at least one discretized weight of a set of discretized weights to compute an influence that the at least one input feature element of the set of input feature elements of the least one input feature map has on one or more output feature elements of at least one output feature map.

Monte Carlo and quasi-Monte Carlo integration are simple numerical recipes for solving complicated integration problems, such as valuating financial derivatives or synthesizing photorealistic images by light transport simulation. A drawback of a straightforward application of (quasi-)Monte Carlo integration is the relatively slow convergence rate that manifests as high error of Monte Carlo estimators. Neural control variates may be used to reduce error in parametric (quasi-)Monte Carlo integration—providing more accurate solutions in less time. A neural network system has sufficient approximation power for estimating integrals and is efficient to evaluate. The efficiency results from the use of a first neural network that infers the integral of the control variate and using normalizing flows to model a shape of the control variate.

Simulation and validation of neural network systems is provided. In various embodiments, a description of an artificial neural network is read. A directed graph is constructed comprising a plurality of edges and a plurality of nodes, each of the plurality of edges corresponding to a queue and each of the plurality of nodes corresponding to a computing function of the neural network system. A graph state is updated over a plurality of time steps according to the description of the neural network, the graph state being defined by the contents of each of the plurality of queues. Each of a plurality of assertions is tested at each of the plurality of time steps, each of the plurality of assertions being a function of a subset of the graph state. Invalidity of the neural network system is indicated for each violation of one of the plurality of assertions.

The invention discloses a method and device for constructing a multiple neural network system for transformer thermal defect identification, and the method comprises the steps: obtaining each excitation parameter set, a temperature value data set on each monitoring point, and a temperature value data set on each feature point of a transformer under each working condition, constructing a first neural network on the basis of each excitation parameter set and the temperature value data set on each feature point under each working condition, constructing a second neural network on the basis of the temperature value data set on each monitoring point and the temperature value data set on each feature point, and combining the first neural network and the second neural network to obtain a first neural network and a second neural network; and a multi-neural network system is obtained. Through the method and the device provided by the embodiment of the invention, a multiple neural network system can be obtained to realize rapid and high-precision transformer temperature field inversion and fault thermal defect identification.

The invention discloses a photon convolutional neural network system, which belongs to the technical field of optical neural networks and comprises dynamic convolution kernel modules in one-to-one correspondence with convolution layers in a pre-trained convolutional neural network model. The dynamic convolution kernel module comprises a photon synaptic device based on a phase change material, and convolution kernel weights corresponding to convolution layers are stored in the photon synaptic device in a non-volatile manner; before the dynamic convolution kernel module performs convolution operation, according to a dynamic coefficient generated by a coefficient generation model according to the content of input data, the weight of a corresponding photon synaptic device in the dynamic convolution kernel module is dynamically adjusted in a maladjustment-prone mode, so that the dynamic convolution kernel module can flexibly adapt to the change of the input data; according to the invention, coding is carried out not only depending on a non-volatile phase state of a phase change material, and volatile modulation coding is further introduced on the basis, so that the coding expression capability of a dynamic convolution kernel module is greatly enriched; on the basis, the feature extraction capability of the convolution kernel in the photon convolutional neural network system provided by the invention is relatively high.

Systems, apparatus, articles of manufacture, and methods to evaluate video quality based on trained neural networks are disclosed. An example apparatus disclosed herein obtains, using a trained neural network, target features corresponding to a target video, the target features based on one or more layers of the trained neural network. The example apparatus also obtains, using the trained neural network, reference features corresponding to a reference video, the reference features based on the one or more layers of the trained neural network, the reference video associated with the target video. The example apparatus further outputs a quality metric for the target video based on the target features, the reference features, and a set of weights. In some examples, the apparatus optionally outputs an error map for the target video.

Systems and methods are disclosed for image classification of electro-optical images and synthetic aperture radar images using training techniques that can include appearance labeling and triplet mining to train a neural network system. The training data can include image pairs of electro-optical images and synthetic radar aperture images. The training data can include anchor, positive, and negative images. The neural network can be trained using triplet loss and cross-entropy loss. The trained neural network can be used for object classification such as automatic target recognition of aerial images.

A method for processing an input audio signal, comprising conditioning a first neural network system with a representation of the input audio signal to predict a bit-rate reduced representation of a processed input audio signal, the first neural network system being trained to generate a bit-rate reduced representation of a processed version of a given audio signal, wherein the bit-rate reduced representation has a format associated with a pre-defined audio encoding process, conditioning a second neural network system with the bit-rate reduced representation to predict an enhanced representation of the processed audio signal, the second neural network system being trained to generate an enhanced representation of a given a bit-rate reduced audio representation, wherein the bit-rate reduced representation has a format associated with the pre-defined audio encoding process, and transforming the enhanced representation of the processed audio signal into an output audio signal.