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

WO2017010049A1 MEMRISTIVE NEUROMORPHIC CIRCUIT AND METHOD FOR TRAINING THE MEMRISTIVE NEUROMORPHIC CIRCUIT 审中-公开
标题翻译：用于训练神经元电磁场的电磁神经元电路及方法
公开(公告)号：WO2017010049A1
公开(公告)日：2017-01-19
申请号：PCT/JP2016/003001
申请日：2016-06-22
申请人： DENSO CORPORATION , THE REGENTS OF THE UNIVERSITY OF CALIFORNIA
发明人： KATAEVA, Irina , STRUKOV, Dmitri B. , MERRIKH-BAYAT, Farnood
IPC分类号： G06N3/063
CPC分类号： G06N3/084 , G06F11/0721 , G06F11/079 , G06N3/063 , G06N3/088
摘要： A neural network (10) is implemented as a memristive neuromorphic circuit that includes a neuron circuit (112, 114) and a memristive device (113) connected to the neuron circuit (112, 114). An input voltage is sensed at a first terminal of the memristive device (113) during a feedforward operation of the neural network (10). An error voltage is sensed at a second terminal of the memristive device (113) during an error backpropagation operation of the neural network (10). In accordance with a training rule, a desired conductance change for the memristive device (113) is computed based on the sensed input voltage and the sensed error voltage. Then a training voltage is applied to the memristive device (113). Here, the training voltage is proportional to a logarithmic value of the desired conductance change.
摘要翻译：神经网络（10）被实现为包含神经元电路（112,114）和连接到神经元电路（112,114）的忆阻器（113）的忆阻神经元电路。在神经网络（10）的前馈操作期间，在忆阻器件（113）的第一端子处感测输入电压。在神经网络（10）的误差反向传播操作期间，在忆阻器（113）的第二端处感测到误差电压。根据训练规则，基于感测的输入电压和感测到的误差电压来计算忆阻装置（113）的期望的电导变化。然后将训练电压施加到忆阻器（113）。这里，训练电压与期望的电导变化的对数值成比例。

2. 发明申请

WO2017010048A1 MEMRISTIVE NEUROMORPHIC CIRCUIT AND METHOD FOR TRAINING THE MEMRISTIVE NEUROMORPHIC CIRCUIT 审中-公开
标题翻译：用于训练神经元电磁场的电磁神经元电路及方法
公开(公告)号：WO2017010048A1
公开(公告)日：2017-01-19
申请号：PCT/JP2016/003000
申请日：2016-06-22
申请人： DENSO CORPORATION , THE REGENTS OF THE UNIVERSITY OF CALIFORNIA
发明人： KATAEVA, Irina , STRUKOV, Dmitri B. , MERRIKH-BAYAT, Farnood
IPC分类号： G06N3/063
CPC分类号： G06N3/0635 , G06N3/08 , G06N3/084
摘要： A neural network (10) is implemented as a memristive neuromorphic circuit (100) that includes a neuron circuit (112, 114) and a memristive device (113) connected to the neuron circuit (112, 114). A conductance balanced voltage pair is provided for the memristive device (113), where the conductance balanced voltage pair includes a set voltage for increasing the conductance of the memristive device (113) and a reset voltage for decreasing the conductance of the memristive device (113). Either the set voltage and reset voltage, when applied to the memristive device (113), effects a substantially same magnitude conductance change in the memristive device (113) over a predetermined range of conductance of the memristive device (113). The provided voltage pair is stored as a conductance balanced map. A training voltage based on the conductance balanced map is applied to the memristive device (113) to train the neural network (10).
摘要翻译：神经网络（10）被实现为包括神经元电路（112,114）和连接到神经元电路（112,114）的忆阻器件（113）的忆阻神经元电路（100）。为忆阻器（113）提供电导平衡电压对，其中电导平衡电压对包括用于增加忆阻器（113）的电导的设定电压和用于降低忆阻器（113）的电导的复位电压）。当施加到忆阻装置（113）时，设定电压和复位电压在忆阻装置（113）的预定电导范围内在忆阻装置（113）中实质上相同的大小电导变化。提供的电压对作为电导平衡图存储。基于电导平衡图的训练电压被应用于忆阻器（113）以训练神经网络（10）。

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