Walter Scott, Jr. College of Engineering

Graduate Exam Abstract

Janarthanan Thivagara Sarma
M.S. Final
Dec 12, 2019, 2:00 pm - 4:00 pm
C101B Engineering
Pruning and Acceleration of Deep Neural Networks
Abstract: Deep Neural Networks are computational and memory intensive applications. Many network compression solutions has been introduced to deploy large trained models in limited memory and time critical systems. We proposed a new methodology that assigns significance rank to the operations in the inference program and for a given area and operation budget, generate only the important operations to do the inference. Our approach has shown that, In many Classical feed forward classification networks we can maintain almost the same accuracy as the original inference by executing less than half of the operations in the original program. We also proposed a methodology to improve the effective implementation of the output sparse computation, controllable by a threshold variable.
Adviser: Louis-Noël Pouchet
Co-Adviser: N/A
Non-ECE Member: Chuck Anderson
Member 3: Sanjay Rajopadhye
Addional Members: Sudeep Pasricha
Travis Augustine, Janarthanan Sarma, Louis-Noël Pouchet, Gabriel Rodríguez, Generating piecewise-regular code from irregular structures, 40th ACM SIGPLAN Conference on Programming Language Design and Implementation, PLDI, Phoenix, AZ (USA), June 2019
Program of Study:
ECE 560
ECE 561
ECE 554
ECE 520
CS 553
CS 445
GRAD 544