Investigation on the feature activation method in a hierarchical concatenation network

Irwan Ramli,C. Ortega-Sanchez
DOI: https://doi.org/10.1109/IC3INA.2017.8251755
2017-10-01
Abstract:This paper presents two different feature activation methods in a Hierarchical Concatenation Network (HCN). In this experiment the similarity measure was conducted by calculating the proportion of the features of tested and reference patterns in all network's layers. The experiment also implemented the circular shifting of input layers. Results show that HCN-2 without circular shifting provides better results compared to HCN-1. HCN-2 also consumes less processing time than HCN-1.
What problem does this paper attempt to address?