Dynamic Neural Diversification: Path to Computationally Sustainable Neural Networks
This work addresses computational sustainability for resource-efficient neural networks, offering incremental improvements in training efficiency and accuracy.
The paper tackles the problem of neuron redundancy in small neural networks, which leads to sub-optimal accuracy or extra training steps, by introducing dynamic decorrelation techniques and a novel weight initialization method that improves early learning and test accuracy by about 40% in the first 5 epochs.
Small neural networks with a constrained number of trainable parameters, can be suitable resource-efficient candidates for many simple tasks, where now excessively large models are used. However, such models face several problems during the learning process, mainly due to the redundancy of the individual neurons, which results in sub-optimal accuracy or the need for additional training steps. Here, we explore the diversity of the neurons within the hidden layer during the learning process, and analyze how the diversity of the neurons affects predictions of the model. As following, we introduce several techniques to dynamically reinforce diversity between neurons during the training. These decorrelation techniques improve learning at early stages and occasionally help to overcome local minima faster. Additionally, we describe novel weight initialization method to obtain decorrelated, yet stochastic weight initialization for a fast and efficient neural network training. Decorrelated weight initialization in our case shows about 40% relative increase in test accuracy during the first 5 epochs.