On randomization of neural networks as a form of post-learning strategy
This addresses the issue of suboptimal training outcomes for practitioners using neural networks, though it appears incremental as it builds on existing post-learning methods.
The paper tackles the local minimum problem in neural network training by proposing a post-learning strategy inspired by quantum effects to search for improved weight configurations with minimal extra computational cost, validated through several numerical experiments.
Today artificial neural networks are applied in various fields - engineering, data analysis, robotics. While they represent a successful tool for a variety of relevant applications, mathematically speaking they are still far from being conclusive. In particular, they suffer from being unable to find the best configuration possible during the training process (local minimum problem). In this paper, we focus on this issue and suggest a simple, but effective, post-learning strategy to allow the search for improved set of weights at a relatively small extra computational cost. Therefore, we introduce a novel technique based on analogy with quantum effects occurring in nature as a way to improve (and sometimes overcome) this problem. Several numerical experiments are presented to validate the approach.