LEARNING OF CONTEXTUAL BANDITS USING ADAPTIVE NEURO FUZZY INFERENCE SYSTEM

Abstract

In the Traditional computer science, we program the computer to achieve some task. In this paradigm we give instructions to a computer to do its task successfully, but now a days we are moving towards a new paradigm which is called machine learning. In this paradigm we provide some labelled examples to a machine, with the help of which it will automatically derive the rules and patterns and save the extracted information to predict the testing data i.e. machine learns to accomplish the task based on the examples provided by us. For example, in Natural Language Processing we give the two large corpus of documents to the machine as an example and the machine will learn to discover patterns in order to match the right words and right expression to go from one language to another. If we try to write the rules for the same then we have to write immunes amount of code and we may not be able to write all the translations form one language to another. In this report, we will illustrate the reinforcement learning using ANFIS (Adaptive Neuro Fuzzy Inference System). The different number of datasets are used as an experiment in the purposed model which is Contextual Multi-Armed Bandit Algorithm named as Adaptive Neuro Fuzzy Inference System (ANFIS). UCI which is a machine learning repository provided the datasets related to these kinds of problems i.e. these datasets comes under the category of CMAB problem.