Local Feature Extraction based KELM for Face Recognition

In neural networks, the success of a learning algorithm used for solving a regression or classification problem is limited by the feature extraction method used. The performance of a classifier depends on the representation of input data. This paper presents entropy based kernel-extreme learning machine (EK-ELM) algorithm for classification problem related to face images. It is a non-iterative, deterministic, single-layer feed-forward neural network learning algorithm, based on the concept of Shannon's entropy and extreme learning machine. In proposed algorithm, entropy is used for the obtaining the local features of an image that are integrated for classification. EK-ELM compares favorably with ELM together with Kernel ELM, on two popular benchmark face datasets. The comparative study formed on the experimental results shows the effectiveness of proposed methodology.