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. 发明申请

WO2018023832A1 SYSTEMS, METHODS AND DEVICES FOR NEURAL NETWORK COMMUNICATIONS 审中-公开
标题翻译：用于神经网络通信的系统，方法和设备
公开(公告)号：WO2018023832A1
公开(公告)日：2018-02-08
申请号：PCT/CN2016/094914
申请日：2016-08-12
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： DALTON, Barnaby , COURVILLE, Vanessa , SALDANA, Manuel
IPC分类号： G06N3/08
CPC分类号： G06N3/08 , G06N3/04 , G06N3/0454 , G06N3/063 , G06N3/084
摘要： A system for training a neural network includes a first set of neural network units and a second set of neural networking units. Each neural network unit in the first set is configured to compute parameter update data for one of a plurality of instances of a first portion of the neural network. Each neural network unit in the first set includes a communication interface for communicating its parameter update data for combination with parameter update data from another neural network unit in the first set. Each neural network unit in the second set is configured to compute parameter update data for one of a plurality of instances of a second portion of the neural network. Each neural network unit in the second set includes a communication interface for communicating its parameter update data for combination with parameter update data from another neural network unit in the second set.
摘要翻译：用于训练神经网络的系统包括第一组神经网络单元和第二组神经网络单元。第一组中的每个神经网络单元被配置为计算神经网络的第一部分的多个实例之一的参数更新数据。第一组中的每个神经网络单元包括通信接口，用于传送其参数更新数据以与来自第一组中的另一个神经网络单元的参数更新数据组合。第二组中的每个神经网络单元被配置为计算神经网络的第二部分的多个实例之一的参数更新数据。第二组中的每个神经网络单元包括通信接口，用于传送其参数更新数据以与来自第二组中另一个神经网络单元的参数更新数据组合。

2. 发明申请

WO2017198168A2 REDUCTION OF PARAMETERS IN FULLY CONNECTED LAYERS OF NEURAL NETWORKS BY LOW RANK FACTORIZATIONS 审中-公开
标题翻译：低秩分解降低神经网络完全连通层的参数
公开(公告)号：WO2017198168A2
公开(公告)日：2017-11-23
申请号：PCT/CN2017/084665
申请日：2017-05-17
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： SOZUBEK, Serdar , DALTON, Barnaby , COURVILLE, Vanessa , TAYLOR, Graham
IPC分类号： G06N3/02
CPC分类号： G06N3/04 , G06F17/16 , G06N3/08
摘要： The present disclosure is drawn to the reduction of parameters in fully connected layers of neural networks. For a layer whose output is defined by y ＝ W x , where y ∈ R m is the output vector, x ∈ R n is the input vector, and W ∈ R mxn is a matrix of connection parameters, matrices U ij and V ij are defined and submatrices W ij are computed as the product of U ij and V ij , so that W ij ＝ V ij U ij , and W is obtained by appending submatrices W i,j .
摘要翻译：本公开涉及神经网络的完全连接层中的参数的减少。对于其输出由 y <=> i _{x 定义的图层，其中< 如果输出向量是输出向量，则X∈R 是输出向量， / i> n 是输入向量，并且W∈R mxn i> 是连接参数的矩阵，矩阵U ij 和< / i> _{ij 是定义的并且子矩阵 ij sub>＆lt; i＆gt;作为＆lt; i＆gt; i＆lt; i＆gt;＆lt; i＆gt;＆lt; i＆gt; i> = V ij U ij i>，和是通过附加子矩阵 W _{i，j 获得的。 >}}}

3. 发明申请

WO2016062251A1 PARALLEL HISTORY SEARCH AND ENCODING FOR DICTIONARY-BASED COMPRESSION 审中-公开
标题翻译：并行历史搜索和编码基于词汇的压缩
公开(公告)号：WO2016062251A1
公开(公告)日：2016-04-28
申请号：PCT/CN2015/092447
申请日：2015-10-21
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： DALTON, Barnaby , LAK, Zahra , MOODY, Brian , SALDANA, Manuel
IPC分类号： G06F17/30
CPC分类号： H03M7/42 , G06F3/0608 , G06F3/0644 , G06F3/0683 , H03M7/30 , H03M7/3084 , H03M7/3088
摘要： Systems, devices and methods for data compression using history search for dictionary based compression. Systems, devices and methods may use parallel processing techniques for data compression and encoding. Systems, devices and methods may provide memory search techniques for hardware.
摘要翻译：使用基于字典的压缩的历史搜索的数据压缩的系统，设备和方法。系统，设备和方法可以使用用于数据压缩和编码的并行处理技术。系统，设备和方法可以为硬件提供存储器搜索技术。

4. 发明申请

WO2017185416A1 SYSTEMS, METHODS AND DEVICES FOR A MULTISTAGE SEQUENTIAL DATA PROCESS 审中-公开
标题翻译：用于多级连续数据处理的系统，方法和设备
公开(公告)号：WO2017185416A1
公开(公告)日：2017-11-02
申请号：PCT/CN2016/081872
申请日：2016-05-12
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： COURVILLE, Vanessa , SALDANA, Manuel , DALTON, Barnaby
IPC分类号： G06T1/20
CPC分类号： G06F3/0656 , G06F3/0608 , G06F3/0673 , G06N3/0454 , G06N3/084
摘要： A circuit for a multistage sequential data process includes a plurality of memory units. Each memory unit is associated with a stage of the sequential data process which, for each data set inputted to the stage, the stage provides an intermediate data set for storage in the associated memory unit for use in at least one subsequent stage of the sequential data process, where each of the plurality of memory units is sized based on relative locations of the stage providing the intermediate data set and the at least one subsequent stage in the sequential data process.
摘要翻译：用于多级时序数据处理的电路包括多个存储单元。每个存储器单元与顺序数据处理的阶段相关联，对于输入到阶段的每个数据集，阶段提供中间数据集以存储在关联的存储器单元中以用于顺序数据的至少一个后续阶段过程，其中基于在顺序数据处理中提供中间数据集和至少一个后续阶段的阶段的相对位置来确定多个存储器单元中的每一个的大小。

5. 发明申请

WO2016062252A1 PARALLEL DICTIONARY-BASED COMPRESSION ENCODER 审中-公开
标题翻译：并行字典基于压缩编码器
公开(公告)号：WO2016062252A1
公开(公告)日：2016-04-28
申请号：PCT/CN2015/092448
申请日：2015-10-21
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： DALTON, Barnaby , LAK, Zahra , SALDANA, Manuel , MOODY, Brian
IPC分类号： H03M7/30 , H04L29/06
CPC分类号： H03M7/42 , G06F3/0608 , G06F3/0644 , G06F3/0683 , H03M7/30 , H03M7/3084 , H03M7/3088
摘要： Systems, devices and methods for data compression using history search for dictionary based compression. Systems, devices and methods may use parallel processing techniques for data compression and encoding. Systems, devices and methods may provide memory search techniques for hardware.
摘要翻译：使用基于字典的压缩的历史搜索的数据压缩的系统，设备和方法。系统，设备和方法可以使用用于数据压缩和编码的并行处理技术。系统，设备和方法可以为硬件提供存储器搜索技术。

6. 发明申请

WO2017198168A3 REDUCTION OF PARAMETERS IN FULLY CONNECTED LAYERS OF NEURAL NETWORKS BY LOW RANK FACTORIZATIONS 审中-公开
公开(公告)号：WO2017198168A3
公开(公告)日：2017-11-23
申请号：PCT/CN2017/084665
申请日：2017-05-17
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： SOZUBEK, Serdar , DALTON, Barnaby , COURVILLE, Vanessa , TAYLOR, Graham
IPC分类号： G06N3/02
摘要： The present disclosure is drawn to the reduction of parameters in fully connected layers of neural networks. For a layer whose output is defined by y ＝ W x , where y ∈ R m is the output vector, x ∈ R n is the input vector, and W ∈ R mxn is a matrix of connection parameters, matrices U ij and V ij are defined and submatrices W ij are computed as the product of U ij and V ij , so that W ij ＝ V ij U ij , and W is obtained by appending submatrices W i,j .

7. 发明申请

WO2016062254A1 MEMORY-BASED HISTORY SEARCH 审中-公开
标题翻译：基于记忆的历史搜索
公开(公告)号：WO2016062254A1
公开(公告)日：2016-04-28
申请号：PCT/CN2015/092453
申请日：2015-10-21
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： LAK, Zahra , SALDANA, Manuel , DALTON, Barnaby
IPC分类号： H04N7/12
CPC分类号： H03M7/42 , G06F3/0608 , G06F3/0644 , G06F3/0683 , H03M7/30 , H03M7/3084 , H03M7/3088
摘要： Systems, devices and methods for data compression using history search for dictionary based compression. Systems, devices and methods may use parallel processing techniques for data compression and encoding. Systems, devices and methods may provide memory search techniques for hardware.
摘要翻译：使用基于字典的压缩的历史搜索的数据压缩的系统，设备和方法。系统，设备和方法可以使用用于数据压缩和编码的并行处理技术。系统，设备和方法可以为硬件提供存储器搜索技术。

8. 发明申请

WO2018192200A1 SYSTEM AND METHOD FOR TRAINING NEURAL NETWORK 审中-公开
公开(公告)号：WO2018192200A1
公开(公告)日：2018-10-25
申请号：PCT/CN2017/106363
申请日：2017-10-16
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： KHARAGHANI, Sepideh , COURVILLE, Vanessa , DALTON, Barnaby
IPC分类号： G06N3/08
CPC分类号： G06N3/082 , G06N3/0472 , G06N3/084
摘要： A method for training a neural network comprising at least one layer comprising a plurality of input nodes, a plurality of output nodes, and a plurality of connections for connecting each one of the plurality of input nodes to each one of the plurality of output nodes, is provided. The method comprises pseudo-randomly selecting a subset of the plurality of connections, each connection of the plurality of connections having associated therewith a weight parameter and a probability of being retained in the neural network, generating output data by feeding input data over the subset of connections, computing an error between the generated output data and desired output data, and, for at least one connection in the subset of connections, determining a contribution of the weight parameter to the error and updating the probability of being retained in the neural network accordingly.

9. 发明申请

WO2018010244A1 SYSTEMS, METHODS AND DEVICES FOR DATA QUANTIZATION 审中-公开
标题翻译：用于数据量化的系统，方法和设备
公开(公告)号：WO2018010244A1
公开(公告)日：2018-01-18
申请号：PCT/CN2016/094915
申请日：2016-08-12
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： SALDANA, Manuel , DALTON, Barnaby , COURVILLE, Vanessa
IPC分类号： G06N3/02 , G06N3/08
摘要： A memory control unit for handling data stored in a memory device includes a first interface to an interconnection with at least one memory bank; a second interface for communicating with a data requesting unit; and a memory quantization unit. The memory quantization unit is configured to: obtain, via the first interface, a first weight value from the at least one memory bank; quantize the first weight value to generate at least one quantized weight value having a shorter bit length than a bit length of the first weight value; and communicate the at least one quantized weight value to the data requesting unit via the second interface.
摘要翻译：用于处理存储在存储器装置中的数据的存储器控制单元包括：与至少一个存储体相互连接的第一接口; 第二接口，用于与数据请求单元进行通信; 和一个存储器量化单元。所述存储器量化单元被配置为：经由所述第一接口从所述至少一个存储体获得第一权重值; 量化所述第一权重值以生成具有比所述第一权重值的比特长度短的比特长度的至少一个量化权重值; 并且通过第二接口将至少一个量化权重值传送给数据请求单元。

10. 发明申请

WO2017198169A1 REDUCTION OF PARAMETERS IN FULLY CONNECTED LAYERS OF NEURAL NETWORKS 审中-公开
标题翻译：参数在完全连通的神经网络中的减少
公开(公告)号：WO2017198169A1
公开(公告)日：2017-11-23
申请号：PCT/CN2017/084666
申请日：2017-05-17
申请人： HUAWEI TECHNOLOGIES CO., LTD.
发明人： DALTON, Barnaby , SOZUBEK, Serdar , SALDANA, Manuel , COURVILLE, Vanessa
IPC分类号： G06F17/30
CPC分类号： G06N3/04 , G06N3/08
摘要： Disclosed is the reduction of parameters in fully connected layers of neural networks. For a layer whose output is defined by y=Wx, where y is the output vector, x is the input vector, and W is a matrix of connection parameters, vectors u ij and v ij are defined and submatrices W i,j are computed as the outer product of u ij and v ij , so that W i,j = v ij ⊗u ij , and W is obtained by appending submatrices W i,j .
摘要翻译：公开了完全连接的神经网络层中参数的减少。对于其输出由y = Wx定义的层，其中y是输出向量，x是输入向量，并且W是连接参数的矩阵，向量u ij i和v ij 定义子矩阵W i，j作为u ij和v ij的外积，从而W 并且W是通过附加子矩阵W i，j，j来获得的。这些子矩阵W i，j，i，j， / p>

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