2019 International Conference on Software, Telecommunications and Computer Networks (SoftCOM) | 2019
Classification using Discriminative Restricted Boltzmann Machines on Spark
Abstract
Apache Spark is a popular open-source framework for distributed execution of data analytics and machine learning tasks, but with limited support for neural networks. In this work, we investigate the reasons for that and based on our findings we extend the native MLLib library with Restricted Boltzmann Machines, a common building block of deep learning architectures, along with a gradient descent optimization procedure suitable for neural networks. As a case study, we developed a handwritten digit recognition application using discriminative RBMs to demonstrate the performance of our implementation.