Online Sequential Extreme Learning algorithm with kernels for bigdata classification

Extreme machine learning and its variants have shown good generalization performance and high leaning speed in many applications through its fast convergence. Despite the parallel and distributed ELM on MapReduce framework able to handle very large scale dataset for bigdata applications, the process of coping up with the rapidly updating data is a challenging one. Among the unified algorithms, the ELM with kernel uses kernels instead of random feature mappings. After, analyzing the property of ELM, it is observed that its most expensive computational part is matrix multiplication and matrix inversion. With the exponentially increasing volume of data, the matrix operations cannot be directly implemented on MapReduce. This paper proposes a novel approach, online sequential ELM with kernel (OS-ELM-Ker) based on sparsification criteria. Consequently, the efficient learning of frequently updated massive dataset is obtained. Based on the extensive experiments on synthetic dataset, it is observed that the proposed OS-ELM-Ker is highly efficient in classifying massive rapidly updated training dataset in terms of generalizations error and training time.