Evolutionary Techniques in Making Efficient Deep-Learning Framework: A Review

Deep learning (DL) which is an important part of artificial intelligence has become an advanced tool for learning processes viz. classification, regression, image processing and natural language processing. While DL has demonstrated notable successes in addressing various real-world challenges in recent years, its performance can be significantly impacted by several factors, including the chosen model architecture and hyperparameters. Evolutionary algorithms can be employed at various stages of the DL framework to enhance the model performance and optimize the hyperparameters. Evolutionary algorithms and metaheuristics are the search techniques inspired by the natural phenomenon and biological evolution which are easy to use because of their non-dependency to mathematical constraints. The present study provides the brief survey of the various evolutionary algorithms employed to the DL framework at different stages to make the model efficient.