首页> 外文会议>International Conference on Holography and Correlation Optics >Lines of optoelectronic neural elements with optical inputs/outputs based on BISPIN-devices for optical neural networks
【24h】

Lines of optoelectronic neural elements with optical inputs/outputs based on BISPIN-devices for optical neural networks

机译:基于光学神经网络的BISPIN设备的带有光学输入/输出的光电神经元线

获取原文

摘要

Abstract: The necessity of multichannel (1D or 2D) neural modules with optical inputs/outputs and completely parallel information processing, carrying out the operations over continuous variables, is shown. These neuromodules have to be multifunctional with programmable tuning. The realization of a definite function tuning is carried out by means of control signals. It is suitable to the functioning description of neural networks based on frequency-dynamic models to use the apparatus of continuous (neuron) logics. The operations of array neuron logic are discussed and it is shown that all these operations can be expressed by means of 2 basic operations: limited difference and addition. The most promising neuron models are the frequency-dynamic ones, because their functioning principles correspond to natural neuron ones. The circuit of optoelectronic cell, which fulfills the limited difference operation, storage, and threshold control operations is proposed. Experimental results for this cell are given. Circuit and experimental parameters for comparator of optical signals are also given. Creation opportunity and promising application of such circuits and lines on its base for optical neural nets are shown. !17
机译:摘要:显示了具有光学输入/输出和完全并行信息处理,对连续变量进行运算的多通道(1D或2D)神经模块的必要性。这些神经模块必须具有可编程调整的多功能性。确定功能调整的实现是通过控制信号实现的。使用连续(神经元)逻辑的设备适合于基于频率动力学模型的神经网络的功能描述。讨论了数组神经元逻辑的操作,并证明所有这些操作都可以通过2种基本操作来表示:有限的差异和加法。最有前途的神经元模型是频率动力学模型,因为它们的功能原理与自然神经元模型相对应。提出了一种光电池单元电路,该电路可以完成有限的差分运算,存储和阈值控制操作。给出了该电池的实验结果。还给出了光信号比较器的电路和实验参数。展示了这种电路和线路在光学神经网络的基础上的创造机会和有希望的应用。 !17

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号