Novel Artificial Immune Networks-based optimization of shallow machine learning (ML) classifiers

摘要: Abstract Artificial Immune Networks (AIN) is a population-based evolutionary algorithm that inspired by theoretical immunology. It applies ideas and metaphors from the biological immune system to solve multi-disciplinary problems. This paper presents novel application of AIN for optimizing shallow machine learning (ML) classification algorithms. accomplishes this task searching best hyper-parameter set specific (also termed model selection), which minimizes training error enhances generalization capability algorithm. We present convergence analysis proposed employ it in conjunction with selected, well-known ML classifiers, namely, an extreme (ELM), support vector (SVM) echo state network (ESN). The performance evaluated terms accuracy time, using range benchmark datasets, compared against grid search as well strategy (ES)-based optimization techniques. An empirical study different datasets demonstrates improved SVM, 2% 5%, ESN 3% 6%, whereas case ELM 9%. Comparative simulation results demonstrate potential alternative optimizer