The winter course on machine intelligence and brain research offers credits for selected IIT M students who are undertaking this course. The course is available for all students from any department in IIT M. The selection criteria for this course is based on the (a)current CGPA of the student (b) Interest in brain research and machine intelligence as demonstrated by a small write up (c) Potential to do a project under the Faculty and Staff of CCBR.

This is a short duration course and invitations for expressions of interest will be open soon following which selected students based on the selection criteria will be offered an offer of acceptance. Please note, due to the short duration of the course, once accepted, withdrawl from the course is only possible within 2 days of acceptance. Failiure to attend the lectures or completion of assignemnts and evaluation means a "U" or "W" grade

Objectives:

To understand the fundamentals of research areas, where Neuroscience, Machine learning and Engineering interact such as Vision, Audition, Natural Language and Reinforcement learning.

This interdisciplinary course will consist of lectures and hands on tutorials and is based on the premise that a two-way interchange between neuroscience and machine learning/AI will be mutually productive. The Lectures will be conducted by the faculty of IIT Madras, interacting with scientists at the Center for Computational Brain Research as well as world renowned international experts visiting the center during the course. The course focuses on four main areas of robust interactions between engineering and neuroscience (Vision, Speech/Audition, Natural Language and Reinforcement Learning) and also includes related topics (classical Machine Learning, Hardware and Statistical Physics).

Course Contents:

Module 1: Vision

• ∙Classical Computer/Machine vision:

  • •Feature Extraction: Canny, Corners - Harris & Hessian Affine;

  • •Image Segmentation: Graph-Cut;

  • •Motion Analysis: - Tracking (MOG), Optical Flow.

  • ∙Deep learning techniques for:

  • •Object recognition and detection,

  • •Depth estimation,

  • •Video based activity recognition.

Module 2: Audition

• ∙The Neurobiology of Speech and Hearing,

• ∙Machine learning in Audition.

Module 3: Reinforcement Learning (RL)

• ∙Reinforcement Learning: Rewards and returns,

• ∙Markov decision process,

• ∙Dynamic programming,

• ∙Temporal difference learning,

• ∙Function approximation, policy gradient approach.

Module 4: Natural Language Processing (NLP)

• ∙Challenges in NLP,An intuitive overview of the role of Machine Learning (ML) in NLP tasks like:

  • •Concept-level Information Retrieval,

  • •Word Sense Disambiguation (WSD),

  • •Probabilistic Parsing,

  • •Machine Translation.

Additional Topics: Topics in ML

• ∙Singular Value Decomposition of matrices encoding the Pointwise Mutual Information between words,

• ∙Skip gram models for learning word representations,

• ∙Global Vectors for words,

• ∙Studying the equivalence between SVD and the newer neural network based methods.

Neurobiology

∙Neuroanatomy of the brain: organizational principles, development and evolution;

• ∙Microcircuits and Mesocircuits;

• ∙Action potentials, synapses, neurotransmitters and modulators.