arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1504.03065math.SP
收藏

神经网络的特征值:超立方体间的插值

Eigenvalues of neutral networks: interpolating between hypercubes

T. Reeves, R. S. Farr, J. Blundell, A. Gallagher, T. M. A. Fink

摘要:

神经网络是汉明图的子图,其主特征值决定了其鲁棒性:在其上进化的种群承受错误的能力。在这里,我们考虑最强大的小型神经网络:在连续维度的超立方体图之间逐点插值的图(正方形中的点,线,线和点,正方形,正方形和立方体中的点,等等)。我们证明了这类图的邻接矩阵的主特征值是以顶点数的对数为界的,并对字母长度大于2的Hamming图猜想了一个类似的结果.

英文摘要:

A neutral network is a subgraph of a Hamming graph, and its principal eigenvalue determines its robustness: the ability of a population evolving on it to withstand errors. Here we consider the most robust small neutral networks: the graphs that interpolate pointwise between hypercube graphs of consecutive dimension (the point, line, line and point in the square, square, square and point in the cube, and so on). We prove that the principal eigenvalue of the adjacency matrix of these graphs is bounded by the logarithm of the number of vertices, and we conjecture an analogous result for Hamming graphs of alphabet size greater than two.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑