公开(公告)号：US20240118582A1

公开(公告)日：2024-04-11

申请号：US17963027

申请日：2022-10-10

申请人： HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP

发明人： YIWEI PENG ,  Yuan YUAN ,  Stanley Cheung

IPC分类号： G02F1/21 ,  G02F1/225 ,  G06N3/04 ,  G06N3/067

CPC分类号： G02F1/212 ,  G02F1/2257 ,  G06N3/0481 ,  G06N3/0675 ,  G02F2203/15

摘要： Systems, devices, and methods are provided for all-optical reconfigurable activation devices for realizing various activations functions using low input optical power. The device and systems disclosed herein include a directional coupler comprising a first phase-shift mechanism and an interferometer coupled to the directional coupler. The interferometer comprises at least one microring resonator and a second phase-shift mechanism coupled to thereto. The interferometer and the directional coupler comprise waveguides formed of a first material, while the microring resonator comprises a waveguide formed of a second material and a third phase-shift mechanism. The second material is provided as a low-loss material having a high Kerr effect and large bandgaps, to generate various nonlinear activation functions. The first, second, and third phase-shift mechanisms are configured to control biases within the disclosed systems and devices to achieve a desired activation function.

公开(公告)号：US20240029357A1

公开(公告)日：2024-01-25

申请号：US17868435

申请日：2022-07-19

申请人： Himanshu ARORA ,  Joshua John HORACSEK ,  Changqing ZOU ,  Wallace Michel Pinto LIRA

发明人： Himanshu ARORA ,  Joshua John HORACSEK ,  Changqing ZOU ,  Wallace Michel Pinto LIRA

IPC分类号： G06T17/20 ,  G06T15/04 ,  G06N3/04

CPC分类号： G06T17/20 ,  G06T15/04 ,  G06N3/0481

摘要： An apparatus, method, and computer-readable medium for rendering a 3D image using a neural network for implicit representation of the image. UV coordinates of a texture map corresponding to a point on the image are calculated using a texture neural network. The image is rendered by applying a color value located at the UV coordinates of the texture map to the point.

公开(公告)号：US20230334290A1

公开(公告)日：2023-10-19

申请号：US17720212

申请日：2022-04-13

申请人： Microsoft Technology Licensing, LLC

发明人： Kiran RAMA

IPC分类号： G06N3/04 ,  G06K9/62

CPC分类号： G06N3/0454 ,  G06N3/0481 ,  G06K9/6256

摘要： Systems and method for deep reinforcement learning are provided. The method includes generating, by a first neural network implemented on a processor, a synthetic data set based on an original data set, providing the original data set and the generated synthetic data set to a second neural network implemented on the processor, generating, by the second neural network, a prediction identifying the original data set and the generated synthetic data set, and based at least in part on the prediction incorrectly identifying the generated synthetic data set, exporting the generated synthetic data set.

公开(公告)号：US20230325642A1

公开(公告)日：2023-10-12

申请号：US17719294

申请日：2022-04-12

申请人： TAIWAN SEMICONDUCTOR MANUFACTURING COMPANY, LTD. ,  NATIONAL TSING HUA UNIVERSITY

发明人： Kea Tiong TANG ,  YuHsiang CHENG

IPC分类号： G06N3/04 ,  G06N3/08

CPC分类号： G06N3/049 ,  G06N3/0481 ,  G06N3/0454 ,  G06N3/08

摘要： A method is provided and includes operations as below: training a spiking neural network (SNN) in a first device to generate multiple first weight values of M bits; calculating multiple second weight values of N bits corresponding to the first weight values according to a threshold value, the number M, and the first weight values, wherein the number N is smaller than the number M; retraining the spiking neural network with the second weight values to update the second weight values; and performing a write operation to save the updated plurality of second weight values in a memory in a second device for performing a spiking neural network operation in the second device.

公开(公告)号：US20230298341A1

公开(公告)日：2023-09-21

申请号：US17696709

申请日：2022-03-16

申请人： Maxar Space LLC

发明人： Manuel Gonzalez-Rivero ,  David R. Herdzik ,  Jonathan C. Harris

IPC分类号： G06V20/13 ,  G06N3/04 ,  G06T1/60

CPC分类号： G06V20/13 ,  G06N3/0481 ,  G06T1/60

摘要： Methods and structures are presented for implementing an automatic target recognition system as a convolutional neural network (CNN) in a satellite or other environment with constrained resources, such as limited memory capacity and limited processing capability. For example, this allows for the automatic target recognition to be implemented on a field programmable gate array (FPGA). Image data is split into subsets of contiguous pixels, with the subsets processed in parallel in a CNN of a corresponding processing node using quantized weight values that are determined in a training process that accounts for the constraints of the automatic target recognition system. The results of the automatic target recognition process is based on the combined output of the processing nodes.

公开(公告)号：US20230267600A1

公开(公告)日：2023-08-24

申请号：US17728882

申请日：2022-04-25

申请人： Samsung Display Co., Ltd.

发明人： Shuhui Qu ,  Qisen Cheng ,  Janghwan Lee

IPC分类号： G06T7/00 ,  G06N3/04

CPC分类号： G06T7/001 ,  G06N3/0454 ,  G06N3/0481 ,  G06T2207/20081 ,  G06T2207/20084

摘要： A system including: a memory, an encoder, a decoder, and a processor, the processor being connected to the memory, the encoder, and the decoder. The system is configured to: receive, at the encoder, an input image, divide, by the encoder, the input image into a plurality of image patches, select, by the encoder, codes corresponding to the plurality of image patches of the input image, from a codebook including the codes. The system is further configured to determine, by the encoder, an assigned code matrix including the codes corresponding to the plurality of image patches of the input image, receive, by the decoder, the assigned code matrix from the encoder. The system is further configured to generate, by the decoder, a reconstructed image based on the assigned code matrix.

公开(公告)号：US20230229910A1

公开(公告)日：2023-07-20

申请号：US17937592

申请日：2022-10-03

申请人： Intel Corporation

发明人： Kevin Brady ,  Sudheendra Kadri ,  Niall Hanrahan

IPC分类号： G06N3/08 ,  G06N3/04 ,  G06F13/28

CPC分类号： G06N3/08 ,  G06N3/0481 ,  G06F13/28 ,  G06F2213/28

摘要： A compute block includes a DMA engine that reads data from an external memory and write the data into a local memory of the compute block. An MAC array in the compute block may use the data to perform convolutions. The external memory may store weights of one or more filters in a memory layout that comprises a sequence of sections for each filter. Each section may correspond to a channel of the filter and may store all the weights in the channel. The DMA engine may convert the memory layout to a different memory layout, which includes a sequence of new sections for each filter. Each new section may include a weight vector that includes a sequence of weights, each of which is from a different channel. The DMA engine may also compress the weights, e.g., by removing zero valued weights, before the conversion of the memory layout.

公开(公告)号：US20230196094A1

公开(公告)日：2023-06-22

申请号：US17560010

申请日：2021-12-22

申请人： INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE

发明人： Chih-Cheng LU ,  Jin-Yu LIN ,  Kai-Cheung JUANG

IPC分类号： G06N3/08 ,  G06N3/04

CPC分类号： G06N3/08 ,  G06N3/0481

摘要： A quantization method for neural network model includes following steps: initializing a weight array of a neural network model, wherein the weight array includes a plurality of initial weights; performing a quantization procedure to generate a quantized weight array according to the weight array, wherein the quantized weight array includes a plurality of quantized weights within a fixed range; performing a training procedure of the neural network model according to the quantized weight array; and determining whether a loss function is convergent in the training procedure and outputting a post-trained quantized weight array when the loss function is convergent.

公开(公告)号：US20230186077A1

公开(公告)日：2023-06-15

申请号：US17841577

申请日：2022-06-15

申请人： NVIDIA CORPORATION

发明人： Hongxu YIN ,  Jan KAUTZ ,  Jose Manuel ALVAREZ LOPEZ ,  Arun MALLYA ,  Pavlo MOLCHANOV ,  Arash VAHDAT

IPC分类号： G06N3/08 ,  G06N3/04

CPC分类号： G06N3/08 ,  G06N3/0481

摘要： One embodiment of the present invention sets forth a technique for executing a transformer neural network. The technique includes computing a first set of halting scores for a first set of tokens that has been input into a first layer of the transformer neural network. The technique also includes determining that a first halting score included in the first set of halting scores exceeds a threshold value. The technique further includes in response to the first halting score exceeding the threshold value, causing a first token that is included in the first set of tokens and is associated with the first halting score not to be processed by one or more layers within the transformer neural network that are subsequent to the first layer.

公开(公告)号：US20230148102A1

公开(公告)日：2023-05-11

申请号：US17721310

申请日：2022-04-14

申请人： Honda Motor Co., Ltd.

发明人： Hengbo MA ,  Chiho CHOI

IPC分类号： H04N19/42 ,  H04N19/172 ,  H04N19/132 ,  H04N19/103 ,  H04N19/50 ,  G06N3/08 ,  G06N3/04

CPC分类号： H04N19/42 ,  H04N19/172 ,  H04N19/132 ,  H04N19/103 ,  H04N19/50 ,  G06N3/08 ,  G06N3/0445 ,  G06N3/0481

摘要： Systems and methods for providing a framework for predicting future frames using diverse sampling are provided. In one embodiment, a method for predicting future frames includes receiving a video having a first frame from a first time and a second frame from a second time. The first frame and the second frame are represented in image space. The method also includes updating a prediction model based on the video. The method further includes determining whether a stopping condition is satisfied. In response to determining that the stopping condition has been satisfied, the method includes generating a plurality of future frames for a third time after the second time. The plurality of future frames is generated based on a normalized distance metric that preserves distance of samples in the latent space to the image space. The method yet further includes selecting a candidate frame from the plurality of future frames.