Fine-Tuning Temperatures in Restricted Boltzmann Machines Using Meta-Heuristic Optimization

Abstract

Restricted Boltzmann Machines (RBM) are stochastic neural networks mainly used for image reconstruction and unsupervised feature learning. An enhanced version, the temperature-based RBM (T-RBM), considers a new temperature parameter during the learning process that influences the neurons' activation. Nevertheless, the major vulnerability of such models concerns selecting an adequate system's temperature, which might lead them to inadequate training or even overfitting when wrongly set, thus limiting the network from predicting or working effectively over unseen data. This paper addresses the problem of selecting a suitable system's temperature through a meta-heuristic optimization process. Meta-heuristic-driven techniques, such as Particle Swarm Optimization, Bat Algorithm, and Artificial Bee Colony are employed to find proper values for the temperature parameter. Additionally, for comparison purposes, three standard temperature values and a random search are used as baselines. The results revealed that optimizing T-RBM is suitable for training purposes, primarily due to their complex fitness landscape, which makes fine-tuning temperatures a non-trivial task.