Multistability and multiperiodicity of neual networks are usually considered in the application of associative memory. In this paper, we study the multistability and multiperiodicity of complexvalued neural networks (CVNNs for short) with one step piecewise linear activation functions. By separating the CVNN into its real and imaginary parts and using state decomposition, we can easily increase the storage capacity by using less neurons. Simulation results are given to illustrative the effectiveness of the theoretical results.
|Number of pages||10|
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Publication status||Published - 2014|
- Complex-valued neural networks