MIMOSNN: Software implementation for MIMO sampling neural network

Abstract

This is a software tool that provides the program implementation for MIMO Sampling Neural Network (SNN) and Wide Learning. MIMO SNN is a novel neural network that combines SISO SNN with SLFNs structure to fill the gap of SISO SNN and trains the neurons with trainable activation functions in SISO SNN instead of weights. Wide Learning is an extended algorithm of MIMO SNN and trains the network by adding and updating the new neurons. The experimental results show that the algorithm has good accuracy and convergence features. The software is beneficial to expand the application prospect of the algorithm in instant training, self-growth, big data, and other aspects.

Keywords

• MIMO
• Trainable activation function
• Wide learning
• Instant training