Design of Synthesizing Multi-valued High-Capacity Auto-associative Memories Based on Complex-Valued Networks
This paper presents a novel design method which is aimed to synthesize arbitrary multi-valued auto-associative memories via complex-valued neural networks. Globally exponential stable criteria are obtained to guarantee that the unique storage prototype can be retrieved. The proposed procedure enables auto-associative memories to be synthesized by satisfying the constraints of inequalities rather than the learning procedure. The main emphasis of the research presented here is on multi-valued high-capacity auto-associative memories via complex-valued networks. The designed auto-associative memories with \((2r+2)^n\) high memory capacities are robust with respect to design parameter selection and extend the scope of application of complex-valued neural networks. The approach of external inputs via complex-valued neural networks avoids spurious equilibria and retrieves the stored patters accurately. Some applicable experiments are given to illustrate the effectiveness and superiority.
KeywordsMulti-valued associative memories Network dynamics Design methods Real-imaginary-type activation External inputs
The authors would like to thank the anonymous referees and editors for their helpful suggestions, which have improved the quality of this paper. This work was supported by National Natural Science Foundation of China (Grant nos. 11571170 and 11501290).
- 4.Kitahara, M., Kobayashi, M.: Fundamental abilities of rotor associative memory. In: 9th IEEE International Conference on Computer and Information Science (ICIS), pp. 497–502 (2010)Google Scholar