公开(公告)号：US20230143371A1

公开(公告)日：2023-05-11

申请号：US17863963

申请日：2022-07-13

申请人： SAMSUNG ELECTRONICS CO., LTD. ,  Seoul National University R&DB Foundation

发明人： HANWOONG JUNG ,  Soonhoi HA ,  Donghyun KANG ,  Duseok KANG

IPC分类号： G06F7/57 ,  G06N3/04

CPC分类号： G06F7/57 ,  G06N3/0481

摘要： A neural network operation apparatus and method are provided. The neural network operation apparatus includes an internal storage configured to store data to perform a neural network operation, an arithmetic logical unit (ALU) configured to perform an operation between the stored data and main data based on an operation control signal, an adder configured to add an output of the ALU and an output of a first multiplexer, wherein the first multiplexer is configured to output one of an output of the adder and the output of the ALU based on a reset signal, a second multiplexer configured to output one of the main data and a quantization result of the stored data based on a phase signal, and a controller configured to control the ALU, the first multiplexer, and the second multiplexer based on the operation control signal, the reset signal, and the phase signal.

公开(公告)号：US20190244105A1

公开(公告)日：2019-08-08

申请号：US15981664

申请日：2018-05-16

申请人： Western Digital Technologies, Inc.

发明人： Luiz M. Franca-Neto

IPC分类号： G06N3/08 ,  G06N3/04

CPC分类号： G06N3/084 ,  G06F15/8046 ,  G06N3/04 ,  G06N3/0481 ,  G06N5/046

摘要： A method of computer processing is disclosed comprising receiving a data packet at a processing node of a neural network, performing a calculation of the data packet at the processing node to create a processed data packet, attaching a tag to the processed data packet, transmitting the processed data packet from the processing node to a receiving node during a systolic pulse, receiving the processed data packet at the receiving node, performing a clockwise convolution on the processed data packet and a counter clockwise convolution on the processed data packet, performing an adding function and backpropagating results of the performed sigmoid function to each of the processing nodes that originally processed the data packet.

公开(公告)号：US20190244079A1

公开(公告)日：2019-08-08

申请号：US16266765

申请日：2019-02-04

申请人： The George Washington University

发明人： Jonathan K. George ,  Volker J. Sorger

IPC分类号： G06N3/04 ,  G06N3/067

CPC分类号： G06N3/049 ,  G06N3/0481 ,  G06N3/0675

摘要： The ability to rapidly identify symmetry and anti-symmetry is an essential attribute of intelligence. Symmetry perception is a central process in human vision and may be key to human 3D visualization. While previous work in understanding neuron symmetry perception has concentrated on the neuron as an integrator, the invention provides the coincidence detecting property of the spiking neuron can be used to reveal symmetry density in spatial data. A synchronized symmetry-identifying spiking artificial neural network enables layering and feedback in the network. The network of the invention can identify symmetry density between sets of data and present a digital logic implementation demonstrating an 8×8 leaky-integrate-and-fire symmetry detector in a field-programmable gate array. The efficiency of spiking neural networks can be harnessed to rapidly identify symmetry in spatial data with applications in image processing, 3D computer vision, and robotics.

公开(公告)号：US20190236436A1

公开(公告)日：2019-08-01

申请号：US16180250

申请日：2018-11-05

申请人： Imagination Technologies Limited

发明人： James Imber ,  Linling Zhang ,  Cagatay Dikici

IPC分类号： G06N3/04 ,  G06F7/499

CPC分类号： G06N3/04 ,  G06F7/4836 ,  G06F7/49942 ,  G06F17/11 ,  G06N3/0472 ,  G06N3/0481 ,  G06N3/063

摘要： Hierarchical methods for selecting fixed point number formats with reduced mantissa bit lengths for representing values input to, and/or output, from, the layers of a DNN. The methods begin with one or more initial fixed point number formats for each layer. The layers are divided into subsets of layers and the mantissa bit lengths of the fixed point number formats are iteratively reduced from the initial fixed point number formats on a per subset basis. If a reduction causes the output error of the DNN to exceed an error threshold, then the reduction is discarded, and no more reductions are made to the layers of the subset. Otherwise a further reduction is made to the fixed point number formats for the layers in that subset. Once no further reductions can be made to any of the subsets the method is repeated for continually increasing numbers of subsets until a predetermined number of layers per subset is achieved.

公开(公告)号：US20190220743A1

公开(公告)日：2019-07-18

申请号：US16247884

申请日：2019-01-15

申请人： IDEMIA IDENTITY & SECURITY FRANCE

发明人： Herve CHABANNE ,  Vincent DESPIEGEL ,  Anouar MELLAKH

IPC分类号： G06N3/08 ,  G06N3/04 ,  G06K9/62

CPC分类号： G06N3/0454 ,  G06N3/0481 ,  G06N3/063

摘要： The present invention relates to a method for learning parameters of a convolutional neural network, CNN, for data classification, the method comprising the implementation, by means for processing data (11) of a server (1), of steps consisting of: (a1) Learning, from an already classified learning database, the parameters of a CNN, called quantized CNN, such that said parameters are valued in a discrete space; (a2) Generating a white-box implementation of at least one layer of said quantized CNN, said white-box implementation being predetermined based on at least one of said learned parameters. The present invention also relates to a method for classifying an input datum.

公开(公告)号：US20190205358A1

公开(公告)日：2019-07-04

申请号：US15857918

申请日：2017-12-29

申请人： Facebook, Inc.

发明人： Abdulkadir Utku Diril ,  Jong Soo Park ,  Nadav Rotem ,  Mikhail Smelyanskiy

IPC分类号： G06F17/16 ,  G06F7/544 ,  G06N3/063

CPC分类号： G06F17/16 ,  G06F7/5443 ,  G06N3/0481 ,  G06N3/063

摘要： A special-purpose, hardware-based accelerator may include an input subsystem configured to receive first and second vectors as operands of a full dot-product operation. The accelerator may also include a sparsity-aware dot-product engine communicatively coupled to the input subsystem and configured to perform adaptive dot-product processing by: (1) identifying, within the first and second vectors, at least one zero-value element and (2) executing, in response to identifying the zero-value element, a reduced dot-product operation that excludes, relative to the full dot-product operation, at least one mathematical operation in which the zero-value element is an operand. The accelerator may also include an output subsystem that is communicatively coupled to the sparsity-aware dot-product engine and configured to send a result of the reduced dot-product operation to a storage subsystem. Various other accelerators, computing systems, and methods are also disclosed.

公开(公告)号：US20190034793A1

公开(公告)日：2019-01-31

申请号：US15658712

申请日：2017-07-25

申请人： LinkedIn Corporation

发明人： Saurabh Kataria ,  Dhruv Arya ,  Ganesh Venkataraman

IPC分类号： G06N3/08 ,  G06F17/30 ,  G06N3/04

CPC分类号： G06N3/08 ,  G06F16/906 ,  G06F16/93 ,  G06F16/9535 ,  G06N3/04 ,  G06N3/0427 ,  G06N3/0454 ,  G06N3/0481 ,  G06N3/084 ,  G06N7/005 ,  G06Q50/10

摘要： In an example embodiment, a machine learning algorithm is used to train a query-based deep semantic similarity neural network to output a query context vector in a vector space that includes both query context vectors and document context vectors. Both the query context vectors and document context vectors are clustered using a clustering algorithm. When an input search query is obtained, the input search query is also passed into the query-based deep semantic similarity neural network and its output document context vector assigned to a first cluster based on the clustering algorithm. Documents within the first cluster are then retrieved in response to the input search query.

公开(公告)号：US20190026550A1

公开(公告)日：2019-01-24

申请号：US15656269

申请日：2017-07-21

申请人： Adobe Systems Incorporated

发明人： Xiao Yang ,  Paul Asente ,  Mehmet Ersin Yumer

IPC分类号： G06K9/00

CPC分类号： G06K9/00456 ,  G06K9/00463 ,  G06K9/3233 ,  G06K9/34 ,  G06K9/6256 ,  G06K9/627 ,  G06K2209/01 ,  G06N3/0454 ,  G06N3/0481 ,  G06N3/08

摘要： Disclosed systems and methods categorize text regions of an electronic document into document object types based on a combination of semantic information and appearance information from the electronic document. A page segmentation application executing on a computing device accesses textual feature representations that represent text portions in a vector space, where a set of pixels from the page is mapped to a textual feature representation. The page segmentation application generates a visual feature representation, which corresponds to an appearance of a document portion including the set of pixels, by applying a neural network to the page of the electronic document. The page segmentation application generates an output page segmentation of the electronic document by applying the neural network to the textual feature representation and the visual feature representation.

公开(公告)号：US20180357531A1

公开(公告)日：2018-12-13

申请号：US15778732

申请日：2016-08-01

申请人： Devanathan GIRIDHARI

发明人： Devanathan GIRIDHARI ,  Singh Sachan DEVENDRA ,  Kumar SHAILESH

IPC分类号： G06N3/04 ,  G06K9/62 ,  G06F15/18

CPC分类号： G06N3/0472 ,  G06F15/18 ,  G06F17/30705 ,  G06K9/6267 ,  G06N3/0481

摘要： A method for text classification and feature selection using class vectors, comprising the steps of receiving a text/training corpus including a plurality of training features representing a plurality of objects from a plurality of classes; learning a vector representation for each of the classes along with word vectors in the same embedding space; training the class vectors and words vectors jointly using skip-gram approach; and performing class vector based scoring for a particular feature; and performing feature selection based on class vectors.

公开(公告)号：US20180336430A1

公开(公告)日：2018-11-22

申请号：US15947009

申请日：2018-04-06

申请人： Denso IT Laboratory, Inc.

发明人： Ikuro Sato ,  Mitsuru Ambai ,  Hiroshi Doi

IPC分类号： G06K9/46 ,  G06T7/10 ,  G06N3/04 ,  G06K9/20

CPC分类号： G06K9/46 ,  G06K9/209 ,  G06K9/6257 ,  G06N3/0454 ,  G06N3/0481 ,  G06N3/063 ,  G06N3/084 ,  G06T7/10 ,  G06T2207/20084

摘要： A recognition system includes: a sensor processing unit (SPU) that performs sensing to output a sensor value; a task-specific unit (TSU) including an object detection part that performs an object detection task based on the sensor value and a semantic segmentation part that performs a semantic segmentation task based on the sensor value; and a generic-feature extraction part (GEU) including a generic neural network disposed between the sensor processing unit and the task-specific unit, the generic neural network being configured to receive the sensor value as an input to extract a generic feature to be input in common into the object detection part and the semantic segmentation part.