Model agnostic quantification of information transfer in deep neural networks

Acces sibility Controls

Implementing Organization

Indian Institute of Technology (IIT)

Principal Investigator

Prof. Sumantra DuttaRoy

Indian Institute Of Technology (IIT) Delhi

CO-Principal Investigator

Dr. Vinayak Abrol

Indraprastha Institute of Information Technology

About

The project aims to create mathematical tools to measure and evaluate the transferability of deep neural network (DNN)-based models for specific tasks. It will use time-frequency and topology-based methods to create an empirically easy-to-compute metric for assessing pre-trained acoustic models. The metric will analyze the trajectory growth of geometric objects passing through DNNs to link expressivity and generalization error in these models. The project also aims to study how these models can integrate spectral and/or temporal task-dependent information. The methods will be analyzed theoretically and empirically for various pre-trained models, downstream tasks, and modalities.

Source

Anusandhan National Research Foundation/Science and Engineering Research Board (SERB), DST 2023-24

Related Research

View All

Funding Organization

Science and Engineering Research Board (SERB), New Delhi

Anusandhan National Research Foundation (ANRF)

Quick Information

Area of Research

Computer Sciences and Information Technology

Start Year

2024

End Year

2027

Sanction Amount

₹ 48.01 L

Status

Ongoing

Contact

sumantra.dutta.roy@gmail.com

Output

No. of Research Paper

Technologies (If Any)

No. of PhD Produced

No. of Patents

Filed : 00

Grant : 00

Acces sibility Controls

Research Projects