Machine Learning and Big Data

13Transfer Learning

Riyanshi Gupta¹, Kartik Krishna Bhardwaj¹ and Deepak Kumar Sharma²*

¹ Department of Instrumentation and Control, Netaji Subhas University of Technology (formerly Netaji Subhas Institute of Technology), New Delhi, India

² Department of Information Technology, Netaji Subhas University of Technology (formerly Netaji Subhas Institute of Technology), New Delhi, India

Abstract

A rapid surge in applicative implementation of machine learning and artificial intelligence has been experienced in recent times. This has lead to a need for improvement in performance as well as resolution of commonly encountered obstacles in the learning paradigms. One such major obstacle faced in the learning and intelligence frameworks is the need for a data collection by an intelligent model to learn the assigned task. The resolution of task learning hurdles has motivated many new schemes in the learning paradigms for intelligent systems. The independence of learning from a prescribed learning data has long been coveted and worked toward, leading to the coinage of reinforcement learning. The idea behind reinforcement learning is inspired by the dynamic learning capacity of the human brain while performing various tasks. This idea is further extended toward learning through a different inference medium, similar in data to be inferred or the inferential model. The capacity of the human brain to use the knowledge gained during a previous task so as to learn to perform a new task, where ...