论文部分内容阅读
神经元网络在图像识别、字符识别、语音识别应用中,取得的成功都是通过硅片实现的。它比第一代神经元信息处理程序更柔软,而且目前的电路集成了“学徒”的功能。 以软件形式评价,神经元网络越来越多地占据了元件世界的主要地位。在硅片上实现该类型的网络不再是研究领域的内容,而是已经成为生产领域的事了。其集成是由很有规律的结构、连续性及其对故障的宽容等有机地构成的。两年多以来,可以说已商品化的神经元软件已经进入第二代,电路特别柔软,尤其是“学徒”功能直接制在集成块上。半导体之间
The success of the neural network in image recognition, character recognition, and speech recognition applications is achieved through silicon. It is softer than the first generation of neuron information processing programs, and the current circuit integrates the functions of an “apprentice”. In the form of software evaluation, neuronal networks increasingly occupy the major position of the component world. Implementing this type of network on silicon is no longer a matter of research, but has become a matter of production. Its integration is organically constituted by a very regular structure, continuity and tolerance to failure. For more than two years, it can be said that commercially available neuron software has entered the second generation, and the circuit is particularly soft, especially the “apprenticeship” function directly on the integrated block. Between semiconductors