专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明公开

EP3404587A1 CNN PROCESSING METHOD AND DEVICE 有权转让
公开(公告)号：EP3404587A1
公开(公告)日：2018-11-21
申请号：EP17738111.8
申请日：2017-01-09
申请人： Tencent Technology (Shenzhen) Company Limited
发明人： ZHU, Ben
IPC分类号： G06N3/06
CPC分类号： G06N3/06 , G06F9/38 , G06N3/04 , G06N3/0454 , G06N3/063
摘要： A CNN processing method and device for use in realizing module computation by a CNN model on an ASIC, and to ensure CNN computational performance and lower computation costs. The method comprising: acquiring an intensity type of a first module in a CNN model (101); if the intensity type of the first module is a computational intensity, deploying the first module in an ASIC and acquiring a plurality of arithmetic device resources of the ASIC occupied by the first module (102); combining the same arithmetic device resources of the plurality of arithmetic device resources of the ASIC occupied by the first module so as to acquire a first combined-resource module (103); and operating the first combined-resource module in the ASIC (104).

2. 发明公开

EP1540908A4 OPTIMIZATION OF MULTI-DIMENSIONAL TIME SERIES PROCESSING FOR SEIZURE WARNING AND PREDICTION 审中-公开
标题翻译：的多维时间序列处理F-优化R中的ATTACK警告与预测
公开(公告)号：EP1540908A4
公开(公告)日：2009-07-01
申请号：EP03796313
申请日：2003-08-27
申请人： UNIV FLORIDA , UNIV ARIZONA
发明人： SACKELLARES JAMES CHRIS , IASEMIDIS LEONIDAS D , SHIAU DENG-SHAN , DANCE LINDA
IPC分类号： A61B5/00 , A61B5/0476 , A61B5/04 , A61B5/048 , A61N1/36 , G06F17/00 , G06N7/00
CPC分类号： A61N1/36082 , A61B5/04008 , A61B5/048 , A61B5/4094 , A61B5/7264 , A61B5/7275 , A61N1/36025 , A61N1/36114 , G06N3/06
摘要： Characterizing the behavior of a chaotic, multi-dimensional system is achieved by measuring each of a number of signals associated with the system, and generating therefrom, a spatio-temporal response based on each signal. Chaoticity profiles are then generated for each spatio-temporal response. Over a period of time, a determination is made as to whether a certain level of dynamic entrainment and/or disentrainment exists between the chaoticity profiles associated with one or more critical channel groups of a selected predictor, where a predictor represents a given number of critical channel groups "x", a given number of channels per group "y", and a total number of channels N. Characterizing the behavior of the system is based on this determination.

3. 发明公开

EP1353293A3 Behavior control apparatus and method 审中-公开
标题翻译：行为控制设备和方法
公开(公告)号：EP1353293A3
公开(公告)日：2008-07-09
申请号：EP03008183.0
申请日：2003-04-08
申请人： HONDA GIKEN KOGYO KABUSHIKI KAISHA
发明人： Ueda, Yugo, K.K. Honda Gijutsu Kenkyusho , Tsujino, Hiroshi, K.K. Honda Gijutsu Kenkyusho
IPC分类号： G06N3/08 , G06N3/04
CPC分类号： G06N3/06 , G06N3/08
摘要： It is objective of the invention to provide a highly reliable control apparatus and method which reduces the amount of calculation required for both learning of an input/output relationship and actual control as well as prevent inappropriate outputs from being generated for inputs which have been never learned. According to one aspect of the invention, pairs of an input pattern vector for learning and a target output are distributed to a class based on the target output. Then, a correspondence between each element of the input pattern vector for learning and the target output is learned only in that class, and a distribution function is calculated for distributing a new input pattern vector for learning to that class. After the completion of this learning, the distribution function is used to determine which class a new input pattern vector detected by a sensor belongs to. Finally, an output is calculated according to the learning result of that class. Therefore, since the range of the outputs corresponding to the inputs may be limited, reliability of the control is improved.

4. 发明公开

EP4425382A1 MODEL TRAINING METHOD AND COMMUNICATION APPARATUS 审中-公开
公开(公告)号：EP4425382A1
公开(公告)日：2024-09-04
申请号：EP22892105.2
申请日：2022-11-11
申请人： Huawei Technologies Co., Ltd.
发明人： YE, Deshi , CHEN, Songyang , XU, Chen , LI, Rong
IPC分类号： G06N3/08
CPC分类号： G06F18/25 , G06N5/04 , G06F18/214 , G06N3/08 , G06N3/06 , G06N3/084 , G06N3/0464
摘要： This application provides a model training method and a communication apparatus. The method includes: An ith device in a kth group of devices performs n*M times of model training, where the ith device completes model parameter exchange with at least one other device in the kth group of devices in every M times of model training, M is a quantity of devices in the kth group of devices, M is greater than or equal to 2, and n is an integer; and the ith device sends a model Mi,n*M to a target device, where the model Mi,n*M is a model obtained by the ith device by completing the n*M times of model training. According to the solutions provided in this application, a convergence speed and utilization of data and computing power in a model training process can be improved, and requirements on computing, storage, and communication capabilities of a device can be reduced.

5. 发明公开

EP3326117A1 SYSTEM AND COMPUTER-BASED METHOD FOR SIMULATING A HUMAN-LIKE CONTROL BEHAVIOUR IN AN ENVIRONMENTAL CONTEXT. 审中-公开
公开(公告)号：EP3326117A1
公开(公告)日：2018-05-30
申请号：EP15742228.8
申请日：2015-07-23
申请人： Toyota Motor Europe , CAMLIN ITALY S.r.l.
发明人： AMBECK-MADSEN, Jonas , RATEAU, Aymeric , MASTROLEO, Marcello , SASSI, Federico , UGOLOTTI, Roberto , BACCHINI, Alessandro , MUSSI, Luca , ASCARI, Luca
IPC分类号： G06N3/06 , G06N3/04 , G06K9/62
CPC分类号： G06N3/004 , G06F17/12 , G06K9/6251 , G06N3/04 , G06N3/06 , G06N99/005 , G10L25/27
摘要： A computer-based method for simulating a human-like decision in an environmental context, comprising: capturing environmental data with at least one sensor, realising a computer based method for realising a bi-directional compression of high dimensional data by compressing the data into a lower-dimensional map, if environmental data are captured during a learning phase of a computer-based model, evaluate the map of compressed data by determining the quality of the map by how well it separates data with different properties, the captured data corresponding to known pre-recorded data that have been pre-evaluated, if environmental data are captured after the learning phase, add new point to the compressed data and generate a signal indicating which human-like decision to use to correspond to the state of the operator.

6. 发明公开

EP2572293A4 NEURAL PROCESSING UNIT 审中-公开
标题翻译： NEURALE VERARBEITUNGSEINHEIT
公开(公告)号：EP2572293A4
公开(公告)日：2013-12-04
申请号：EP11783883
申请日：2011-01-21
申请人： UNIV CALIFORNIA
发明人： PALMER DOUGLAS A , FLOREA MICHAEL
IPC分类号： G06F15/18 , G06F9/06 , G06N3/06
CPC分类号： G06N3/06 , G06N3/063
摘要： The subject matter disclosed herein provides methods, apparatus, and articles of manufacture for neural-based processing. In one aspect, there is provided a method. The method may include reading, from a first memory, context information stored based on at least one connection value; reading, from a second memory, an activation value matching the at least one connection value; sending, by a first processor, the context information and the activation value to at least one of a plurality of microengines to configure the at least one microengine as a neuron; and generating, at the at least one microengine, a value representative of an output of the neuron. Related apparatus, systems, methods, and articles are also described.
摘要翻译：本文公开的主题提供了用于基于神经的处理的方法，装置和制造物品。在一个方面中，提供了一种方法。该方法可以包括：从第一存储器读取基于至少一个连接值存储的上下文信息; 从第二存储器读取与所述至少一个连接值匹配的激活值; 由第一处理器将所述上下文信息和所述激活值发送到多个微引擎中的至少一个，以将所述至少一个微引擎配置为神经元; 以及在所述至少一个微引擎上生成表示所述神经元的输出的值。还描述了相关的装置，系统，方法和物品。

7. 发明公开

EP2602749A1 Neural working memory device and method for operating same 有权
标题翻译： Neurale arbeitende Speichervorrichtung und Verfahren zum Betrieb davon
公开(公告)号：EP2602749A1
公开(公告)日：2013-06-12
申请号：EP12196277.3
申请日：2012-12-10
申请人： Samsung Electronics Co., Ltd
发明人： Lee, Jun Haeng , Shin, Chang Woo , Ryu, Hyun Surk , Park, Keun Joo
IPC分类号： G06N3/04 , G06N3/063
CPC分类号： G06N3/06 , G06N3/049 , G06N3/063
摘要： A spiking neuron-based working memory device is provided. The spiking neuron-based working memory device includes an input interface configured to convert input spike signals into respective burst signals having predetermined forms, and output a sequence of the burst signals, the burst signals corresponding to the input spike signals in a burst structure, and two or more memory elements (MEs) configured to sequentially store input values respectively corresponding to the outputted sequence of the burst signals, each of the MEs continuously outputting spike signals respectively corresponding to the stored features.
摘要翻译：提供了一种基于神经元的工作记忆装置。尖峰神经元工作存储装置包括：输入接口，被配置为将输入尖峰信号转换成具有预定形式的各自的脉冲串信号，并且输出脉冲信号序列，脉冲串信号对应于突发结构中的输入尖峰信号;以及两个或多个存储器元件（ME）被配置为顺序地存储分别对应于所输出的突发信号序列的输入值，每个ME连续地输出对应于存储的特征的尖峰信号。

8. 发明公开

EP4375892A1 DISTRIBUTED TRAINING METHOD FOR AI MODEL AND RELATED DEVICE 审中-实审
公开(公告)号：EP4375892A1
公开(公告)日：2024-05-29
申请号：EP22857667.4
申请日：2022-08-11
申请人： Huawei Cloud Computing Technologies Co., Ltd.
发明人： LIAN, Yunwen , LI, Yi , JIN, Xiaoxian
IPC分类号： G06N20/00
CPC分类号： G06F18/214 , G06N20/00 , G06N3/06 , G06N3/0464
摘要： This application relates to the field of artificial intelligence technologies, and provides a distributed training method for an AI model and a related device. The method is applied to an AI platform, the AI platform is associated with a computing resource pool, the computing resource pool includes a plurality of computing nodes for distributed training of an AI model, and each of the plurality of computing nodes performs a training task for the distributed training of the AI model. The method includes: performing fault isolation on a first computing node, where the first computing node is a faulty computing node in the plurality of computing nodes; determining a second computing node, where the second computing node is a computing node other than the plurality of computing nodes in the computing resource pool; and configuring the second computing node, so that the second computing node replaces the first computing node to execute a training task. In embodiments of this application, fault recovery duration can be shortened.

9. 发明公开

EP2724297A1 METHOD AND APPARATUS FOR A LOCAL COMPETITIVE LEARNING RULE THAT LEADS TO SPARSE CONNECTIVITY 有权
标题翻译：方法和设备的局部WETTBERWERBS学习规则，稀疏连通LEADS
公开(公告)号：EP2724297A1
公开(公告)日：2014-04-30
申请号：EP12730136.4
申请日：2012-06-21
申请人： Qualcomm Incorporated
发明人： APARIN, Vladimir
IPC分类号： G06N3/08
CPC分类号： G06N3/08 , G05B13/024 , G05B13/027 , G06N3/06 , G06N3/088
摘要： Certain aspects of the present disclosure support a local competitive learning rule applied in a computational network that leads to sparse connectivity among processing units of the network. The present disclosure provides a modification to the Oja learning rule, modifying the constraint on the sum of squared weights in the Oja rule. This constraining can be intrinsic and local as opposed to the commonly used multiplicative and subtractive normalizations, which are explicit and require the knowledge of all input weights of a processing unit to update each one of them individually. The presented rule provides convergence to a weight vector that is sparser (i.e., has more zero elements) than the weight vector learned by the original Oja rule. Such sparse connectivity can lead to a higher selectivity of processing units to specific features, and it may require less memory to store the network configuration and less energy to operate it.
摘要翻译：本公开的某些方面支持在计算网络应用的本地竞争学习规则的网络处理单元之间没有导致稀疏的连接。本发明提供的修饰的Oja的学习规则，修改对平方的权重的在Oja的规则之和约束。相对于常用的乘法和减归一化，这是明确的，需要处理单元的所有输入的权重知识单独更新他们中的每一个这种约束可以是内在的和局部的。所呈现的规则提供收敛的权重向量做是稀疏（即，有更多的零种元素）比由原始Oja的规则学习的权重向量。搜索稀疏连接可导致处理单元对特定特征的更高的选择性，并且可能需要较少的存储网络配置和较少的操作它的能量存储器。

10. 发明公开

EP2572293A1 NEURAL PROCESSING UNIT 审中-公开
标题翻译：神经处理单元
公开(公告)号：EP2572293A1
公开(公告)日：2013-03-27
申请号：EP11783883.9
申请日：2011-01-21
申请人： The Regents of The University of California
发明人： PALMER, Douglas, A. , FLOREA, Michael
IPC分类号： G06F15/18 , G06F9/06 , G06N3/06
CPC分类号： G06N3/06 , G06N3/063
摘要： The subject matter disclosed herein provides methods, apparatus, and articles of manufacture for neural-based processing. In one aspect, there is provided a method. The method may include reading, from a first memory, context information stored based on at least one connection value; reading, from a second memory, an activation value matching the at least one connection value; sending, by a first processor, the context information and the activation value to at least one of a plurality of microengines to configure the at least one microengine as a neuron; and generating, at the at least one microengine, a value representative of an output of the neuron. Related apparatus, systems, methods, and articles are also described.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式