Two Novel Performance Improvements for Evolving CNN Topologies
This work addresses the problem of computational efficiency for researchers and practitioners using evolutionary methods to design CNNs, though it is incremental as it builds on existing genetic algorithm frameworks.
The paper tackled the high computational cost of using genetic algorithms to evolve CNN topologies for image recognition by introducing two novel approaches, resulting in a nearly 20% reduction in complexity and training time while maintaining accuracy on the CIFAR10 dataset.
Convolutional Neural Networks (CNNs) are the state-of-the-art algorithms for the processing of images. However the configuration and training of these networks is a complex task requiring deep domain knowledge, experience and much trial and error. Using genetic algorithms, competitive CNN topologies for image recognition can be produced for any specific purpose, however in previous work this has come at high computational cost. In this work two novel approaches are presented to the utilisation of these algorithms, effective in reducing complexity and training time by nearly 20%. This is accomplished via regularisation directly on training time, and the use of partial training to enable early ranking of individual architectures. Both approaches are validated on the benchmark CIFAR10 data set, and maintain accuracy.