A Google Approach for Computational Intelligence in Big Data

With the advent of the emerging field of big data, it is becoming increasingly important to equip machine learning algorithms to cope with volume, variety, and velocity of data. In this work, we employ the MapReduce paradigm to address these issues as an enabling technology for the well-known support vector machine to perform distributed classification of skin segmentation. An open source implementation of MapReduce called Hadoop offers a streaming facility, which allows us to focus on the computational intelligence problem at hand, instead of focusing on the implementation of the learning algorithm. This is the first time that support vector machine has been proposed to operate in a distributed fashion as it is, circumventing the need for long and tedious mathematical derivations. This highlights the main advantages of MapReduce - its generality and distributed computation for machine learning with minimum effort. Simulation results demonstrate the efficacy of MapReduce when distributed classification is performed even when only two machines are involved, and we highlight some of the intricacies of MapReduce in the context of big data.