Weightless Neural Networks Based on Multi-valued Probabilistic Logic for Node for the Handwritten Digit Classification

This work analyzes the impact of the hyper-parameters of a weightless neural network based on Multi-valued Probabilistic Logic Neurons (MPLN) in order to design an efficient and concise network topology. The study is done based on the implementation of several MPLN architectures for the handwritten digit identification application. The analysis is performed by varying one given parameter while the others are kept unchanged. This allows the impact evaluation of such parameter on the classification accuracy, necessary epoch number to train the network and required processing time. The present work further proposes a modification in the MPLN network for multi-class problems, termed the Mod-MPLN network. The Mod-MPLN network is defined by a change in the network training algorithm and by the inclusion of a specific discriminator at the network output, without changing the intrinsic characteristics of the MPLN-based topology.