Signal-To-Noise-Optimal scaling of heterogenous population codes

作     者：Leibold, Christian 

作者机构：Univ Munich Dept Biol 2 D-82152 Planegg Martinsried Germany 

出 版 物：《NETWORK-COMPUTATION IN NEURAL SYSTEMS》 (网络：神经系统计算)

年 卷 期：2013年第24卷第3期

页      面：114-128页

核心收录：

学科分类：0808[工学-电气工程] 1001[医学-基础医学(可授医学、理学学位)] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：information theory network models population coding 

摘      要：Similarity measures for neuronal population responses that are based on scalar products can be little informative if the neurons have different firing statistics. Based on signal-to-noise optimality, this paper derives positive weighting factors for the individual neurons response rates in a heterogeneous neuronal population. The weights only depend on empirical statistics. If firing follows Poisson statistics, the weights can be interpreted as mutual information per spike. The scaling is shown to improve linear separability and clustering as compared to unscaled inputs.