Functional Link Neural Network with Modified Artificial Bee Colony for Data Classification

Abstract

Functional Link Neural Network (FLNN) has emerged as an important tool for solving non-linear classification problem and has been successfully applied in many engineering and scientific problems. The FLNN structure is much more modest than ordinary feed forward network like the Multilayer Perceptron (MLP) due to its flat network architecture which employs less tuneable weights for training. However, the standard Backpropagation (BP) learning uses for FLNN training prone to get trap in local minima which affect the FLNN classification performance. To recover the BP-learning drawback, this paper proposes an Artificial Bee Colony (ABC) optimization with modification on bee foraging behaviour (mABC) as an alternative learning scheme for FLNN. This is motivated by good exploration and exploitation capabilities of searching optimal weight parameters exhibit by ABC algorithm. The result of the classification accuracy made by FLNN with mABC (FLNN-mABC) is compared with the original FLNN architecture with standard Backpropagation (BP) (FLNN-BP) and standard ABC algorithm (FLNN-ABC). The FLNN-mABC algorithm provides better learning scheme for the FLNN network with average overall improvement of 4.29% as compared to FLNN-BP and FLNN-ABC.