Research Theme - Brain Inspired Learning Algorithms

Artificial Intelligence (AI) has undergone a major paradigm shift over the past decade due to the unreasonable effectiveness of deep learning algorithms. The deep learning algorithms include deep architectures like deep neural network, convolutional neural network, recurrent neural network and long short term memory etc. The unreasonability in these learning algorithms is due to the lack of a strong theoretical foundation on the working of deep learning. The term “deep” refers to the complexity of the architecture used in these learning algorithms. These algorithms learn from data i.e, they are “data-driven” techniques and has applications in various disciplines such as speech recognition, object detection, natural language processing which can be seen as some of the milestones of AI. One of the goals of AI is to develop a machine that mimics the human brain and even though deep learning has several applications, it does not learn the way the human brain learns. This signifies a wide gap between learning in algorithms and the human brain. This motivates to propose a brain inspired novel architecture for learning which uses properties such as stochastic resonance, chaotic synchronization, neural interference, multiplexing, neural plasticity, and causal reasoning so as to improve the performance of learning algorithms. The potential applications of the proposed brain inspired novel architecture are in the fields of healthcare and natural language processing.