公开(公告)号：US20190251398A1

公开(公告)日：2019-08-15

申请号：US16275186

申请日：2019-02-13

申请人： Slingshot Aerospace, Inc.

发明人： David Stuart Godwin, IV ,  Thomas Scott Ashman ,  Spencer Ryan Romo ,  Melanie Stricklan ,  Carrie Inez Hernandez

IPC分类号： G06K9/62 ,  G06K9/00 ,  G06N3/08 ,  G06N3/04

CPC分类号： G06K9/6262 ,  G06K9/00657 ,  G06K9/6202 ,  G06K9/6256 ,  G06N3/04 ,  G06N3/0436 ,  G06N3/0454 ,  G06N3/08 ,  G06N3/084

摘要： Method, electronic device, and computer readable medium embodiments are disclosed. In one embodiment, a method includes training a neural network using a first image dataset and a first truth dataset, then using the trained neural network to analyze a second image dataset. The training includes modifying a loss function of the neural network to forego penalizing the neural network when a feature is predicted with higher than a first confidence level by the neural network, and the first truth dataset has no feature corresponding to the predicted feature.

公开(公告)号：US20190244108A1

公开(公告)日：2019-08-08

申请号：US16270681

申请日：2019-02-08

申请人： Cognizant Technology Solutions U.S. Corporation

发明人： Elliot Meyerson ,  Risto Miikkulainen

IPC分类号： G06N3/08 ,  G06N3/04

CPC分类号： G06N3/084 ,  G06N3/04 ,  G06N3/0445 ,  G06N3/0454

摘要： A multi-task (MTL) process is adapted to the single-task learning (STL) case, i.e., when only a single task is available for training. The process is formalized as pseudo-task augmentation (PTA), in which a single task has multiple distinct decoders projecting the output of the shared structure to task predictions. By training the shared structure to solve the same problem in multiple ways, PTA simulates the effect of training towards distinct but closely-related tasks drawn from the same universe. Training dynamics with multiple pseudo-tasks strictly subsumes training with just one, and a class of algorithms is introduced for controlling pseudo-tasks in practice.

公开(公告)号：US20190244107A1

公开(公告)日：2019-08-08

申请号：US16262878

申请日：2019-01-30

申请人： HRL Laboratories, LLC

发明人： Zachary Murez ,  Soheil Kolouri ,  Kyungnam Kim

IPC分类号： G06N3/08 ,  G06N20/00 ,  G06T7/00

CPC分类号： G06N3/084 ,  G06N20/00 ,  G06T7/0002 ,  G06T2207/20081 ,  G06T2207/20084

摘要： Described is a system for adapting a deep convolutional neural network (CNN). A deep CNN is first trained on an annotated source image domain. The deep CNN is adapted to a new target image domain without requiring new annotations by determining domain agnostic features that map from the annotated source image domain and a target image domain to a joint latent space, and using the domain agnostic features to map the joint latent space to annotations for the target image domain.

公开(公告)号：US20190244081A1

公开(公告)日：2019-08-08

申请号：US15981735

申请日：2018-05-16

申请人： Western Digital Technologies, Inc.

发明人： Luiz M. Franca-Neto

IPC分类号： G06N3/04 ,  G06N3/08 ,  G06N5/04

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：US20190244078A1

公开(公告)日：2019-08-08

申请号：US16233968

申请日：2018-12-27

申请人： Western Digital Technologies, Inc.

发明人： Luiz M. Franca-Neto

IPC分类号： G06N3/04 ,  G06N3/08

CPC分类号： G06N3/063 ,  G06K9/00973 ,  G06K9/6215 ,  G06K9/6217 ,  G06K9/6262 ,  G06N3/04 ,  G06N3/0445 ,  G06N3/08 ,  G06N3/084 ,  G06T5/00 ,  G06T2207/20081 ,  G06T2207/20084

摘要： Some embodiments include a special-purpose hardware accelerator that can perform specialized machine learning tasks during both training and inference stages. For example, this hardware accelerator uses a systolic array having a number of data processing units (“DPUs”) that are each connected to a small number of other DPUs in a local region. Data from the many nodes of a neural network is pulsed through these DPUs with associated tags that identify where such data was originated or processed, such that each DPU has knowledge of where incoming data originated and thus is able to compute the data as specified by the architecture of the neural network. These tags enable the systolic neural network engine to perform computations during backpropagation, such that the systolic neural network engine is able to support training.

公开(公告)号：US20190244077A1

公开(公告)日：2019-08-08

申请号：US15981624

申请日：2018-05-16

申请人： Western Digital Technologies, Inc.

发明人： Luiz M. Franca-Neto

IPC分类号： G06N3/04 ,  G06N3/08

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：US20190238138A1

公开(公告)日：2019-08-01

申请号：US16257492

申请日：2019-01-25

申请人： Seiko Epson Corporation

发明人： Yasuhiro SUDO ,  Hideo HANEDA

IPC分类号： H03L1/02 ,  H03L1/04 ,  H03B5/32 ,  H03B5/04 ,  G06N3/08

CPC分类号： H03L1/025 ,  G06N3/084 ,  H03B5/04 ,  H03B5/32 ,  H03L1/04

摘要： An integrated circuit device includes a digital signal processing circuit that generates frequency control data by performing a temperature compensation process by a neural network calculation process based on temperature detection data and an amount of change in time of the temperature detection data, and an oscillation signal generation circuit that generates an oscillation signal of a frequency set by the frequency control data using a resonator.

公开(公告)号：US20190236455A1

公开(公告)日：2019-08-01

申请号：US16264625

申请日：2019-01-31

申请人： ROYAL BANK OF CANADA

发明人： Matthew Edmund TAYLOR ,  Gabriel Victor DE LA CRUZ, JR. ,  Yunshu DU

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

CPC分类号： G06N3/084 ,  G06K9/6267 ,  G06N3/0472

摘要： Disclosed herein are a system and method for providing a machine learning architecture based on monitored demonstrations. The system may include: a non-transitory computer-readable memory storage; at least one processor configured for dynamically training a machine learning architecture for performing one or more sequential tasks, the at least one processor configured to provide: a data receiver for receiving one or more demonstrator data sets, each demonstrator data set including a data structure representing the one or more state-action pairs; a neural network of the machine learning architecture, the neural network including a group of nodes in one or more layers; and a pre-training engine configured for processing the one or more demonstrator data sets to extract one or more features, the extracted one or more features used to pre-train the neural network based on the one or more state-action pairs observed in one or more interactions with the environment.

公开(公告)号：US20190228312A1

公开(公告)日：2019-07-25

申请号：US15880339

申请日：2018-01-25

申请人： SparkCognition, Inc.

发明人： Sari Andoni ,  Kevin Gullikson

IPC分类号： G06N3/08 ,  G06K9/62 ,  G06N3/04 ,  G06F17/18

CPC分类号： G06N3/088 ,  G06F17/18 ,  G06K9/6218 ,  G06N3/0454 ,  G06N3/084 ,  G06N3/086

摘要： During training mode, first input data is provided to a first neural network to generate first output data indicating that the first input data is classified in a first cluster. The first input data includes at least one of a continuous feature or a categorical feature. Second input data is generated and provided to at least one second neural network to generate second output data. The at least one second neural network corresponds to a variational autoencoder. An aggregate loss corresponding to the second output data is determined, including at least one of evaluating a first loss function for the continuous feature or evaluating a second loss function for the categorical feature. Based on the aggregate loss, at least one parameter of at least one neural network is adjusted. During use mode, the neural networks are used to determine cluster identifications and anomaly likelihoods for received data samples.

公开(公告)号：US20190220729A1

公开(公告)日：2019-07-18

申请号：US15870057

申请日：2018-01-12

申请人： ABL IP HOLDING LLC

发明人： Min-Hao Michael Lu ,  Michael Miu ,  Eric J. Johnson

IPC分类号： G06N3/04 ,  G06N3/08 ,  G05B13/02 ,  H05B37/02

CPC分类号： G06N3/04 ,  G05B13/027 ,  G06N3/084 ,  H04B1/02 ,  H04B1/06 ,  H04B17/318 ,  H05B33/0845 ,  H05B37/0227 ,  H05B37/0272

摘要： Disclosed herein is system level occupancy counting in a lighting system configured to obtain an indicator data of a RF spectrum signal (signal) generated at a number of times in an area. At each respective one of the number of times, apply one of a plurality of heurist algorithm heuristic algorithm coefficients to each indicator data of the signal, based on results of the application of the heuristic algorithm coefficients, generate an indicator data metric value for each of the indicator data for the respective time. The lighting system is also configured to process each of the indicator data metric value to compute a plurality of metric values for the respective time and combine the plurality of metric values to compute an output metric value for each of a plurality of probable number of occupants in the area for the respective time. The lighting system is further configured to determine an occupancy count in the area at the respective time based on the computed output metric value.