Lectures by Pratiksha Trust Young Investigators

Date: Tuesday, December 4, 2018
Time: 2 to 4 PM
Venue: Golden Jubilee Hall, Department of ECE

The Pratiksha Trust, founded by Mr. Kris Gopalakrishnan and Mrs. Sudha Gopalakrishnan, has generously endowed Young Investigator awards at the Indian Institute of Science. This event features four talks by the faculty members who received this award in 2017.

Dr. Sridharan Devarajan, Centre for Neuroscience, IISc
Title: Neuro-computational approaches for understanding brain structure and function

Abstract:

Understanding essential principles by which the complex machinery of the brain produces cognition is a fundamental challenge for 21st century science. Emerging approaches at the interface of neuroscience, computer science and engineering provide a powerful paradigm for meeting this challenge. In this talk, I will present two projects in our lab of this flavor. First, I will describe the recent development of a GPU-based algorithm (ReAl-LiFE) which enables efficient discovery of individualized brain connectomes in large databases. In conjunction with machine learning, ReAl-LiFE has helped identify key signatures of pathological connectivity in Alzheimer’s Disease (collaborative project with Prof. Franco Pestilli, Indiana University and Prof. Partha Talukdar, IISc). Second, I will describe our ongoing efforts to understand mechanisms by which attention works in the brain, using an EEG-based real-time brain-computer interface, highlighting key engineering, math and computational challenges along the way (collaborative project with Prof. Byron Yu, Carnegie Mellon University). I will discuss advantages of such interdisciplinary approaches for understanding brain function, in health and in disease.

Dr. Sriram Ganapathy, Department of EE, IISc
Title: Decoding the Speech and Language Code in the Brain

Abstract:

Speech and language are perceived in the brain at multiple regions responding to different components like speaker, content, language etc. This talk will attempt to tease apart the different aspects of speech and language perception. The first part of the study attempts to extract behavioral cues for speaker representations from the brain. This study examines human talker change detection (TCD) in multi-party speech utterances using a novel behavioral paradigm in which listeners indicate the moment of perceived talker change. By relating the reaction time in TCD to acoustic dimensions using a modeling approach, I will illustrate that the human reaction time can be well predicted using the distance between acoustic features before and after change instant. The second part of the talk analyzes speech representations in EEG exhibited in a language learning task. The key question here is to understand how the brain representations in EEG change as we learn new languages. The aim of the study was two fold, (1) to find representation differences in EEG during the learning of words from a new language and (2) to find neural correlates of language learning through EEG. Throughout the talk, I will attempt to make connections with machine side of speech processing.

Dr. Prasanta Kumar Ghosh, Department of EE, IISc
Title: Research in speech production – from scientific understanding to technological solutions for health-care and language learning

Abstract:

In this talk, I shall present a summary of research activities that happened in SPIRE Lab at IISc over last one year in the area of speech production. Key aspects in these activities include analysis and modeling of rich multi-modal data that captures various aspects of human speech production for seeking answers to fundamental questions in speech science. The research activities also include development of engineering solutions for various clinical applications associated to speech production as well as language learning.

Dr. Chetan Singh Thakur, Department of ESE, IISc
Title: Low-Power Neuromorphic Electronic Systems

Abstract:

Due to the proliferation of internet-of-things (IoTs) in the areas of ubiquitous sensing there has been an increased demand towards integrating intelligence directly onto the IoT hardware platform. The machine learning architecture embedded into these platforms needs to be as energy-efficient as possible. In this regard, wake-up systems play an integral role and operate by flagging on the computationally and power-intensive modules only when some ambient conditions are detected. In this talk, our latest work in building low-power wakeup module using CMOS-Memristor based neuromorphic architectures will be presented. Second part of the talk will present the novel approach of object detection and localization in compressive sensed video using pixel-wise coded exposure imager chip.

Coffee/Tea: 4 PM

All Are Welcome