Storage Capacity of a Multilayer Neural Network with Binary Weights
Storage Capacity of a Multilayer Neural Network with Binary Weights
复制标题
具有二进制权重的多层神经网络的存储容量
DOI：
10.1209/0295-5075/14/2/003
复制
发表时间：
1991
期刊：
影响因子：
1.8
通讯作者：
I. Kanter
中科院分区：
文献类型：
--
作者：
E. Barkai;I. Kanter
应用统计力学方法估计具有二进制权重的两层前馈网络的最大单位重量容量(αc)，该网络充当隐层单元的奇偶机。对于K个≥和2个隐含单元，达到了最大理论容量，αc=1，不同解之间的平均重叠为零。这些结果与模拟结果相吻合。在有限温度下，找到了复型对称破缺的一步解，这似乎是精确的。
Statistical mechanics is applied to estimate the maximal capacity per weight (αc) of a two-layer feed-forward network with binary weights, functioning as a parity machine of the hidden units. For K ≥ 2 hidden units, the maximal theoretical capacity is achieved, αc = 1, and the average overlap between different solutions is zero. These results agree with the simulations. At finite temperature one-step replica symmetry breaking solution is found, which appears to be exact.