Modular Approach to Big Data using Neural Networks

Machine learning can be used to recognize patterns, classify data into classes and make predictions. Neural Networks are one of the many machine learning tools that are capable of performing these tasks. The greatest challenges that we face while dealing with the IBM Watson dataset is the high amount of dimensionality, both in terms of the number of features the data has, as well the number of rows of data we are dealing with. The aim of the project is to identify a course of action that can be chosen when dealing with similar problems. The project aims at setting up and experimenting with different strategies of training neural networks in order to reduce training time and increase prediction accuracy. The project will contrast the advantages and disadvantages of using these modular approaches and provide a completely open source implementation of the system