Oct 31, 2023, Tue (day 1)  10am ~ 12pm,  Location: 92383 電機系館三樓三丙教室

Nov 01, 2023 Wed (day 2) 10am ~ 12pm, Location: 92119 電機系館一樓靄雲廳

Dataflow-based Design and Implementation of Embedded Signal

Processing and Machine Learning Systems: A Short Course

National Cheng Kung University, Taiwan, Oct. 31 & Nov. 1, 2023

Instructor: Shuvra S. Bhattacharyya

Professor, University of Maryland at College Park, USA, and

Chair of Excellence, INSA/IETR, Rennes, France,

Email: ssb@umd.edu, Web: http://www.ece.umd.edu/~ssb

Overview

In this short course, we will cover design methodologies for embedded signal processing and machine learning (SPML) systems based on dataflow models of computation. Dataflow modeling provides an important class of design abstractions that helps to address the challenges brought about by the increasing complexity of

embedded SPML applications, and the complexity and heterogeneity of the ardware platforms that are used to deploy these applications. To exploit the potential of state-of-the-art embedded platforms and advanced SPML algorithms, important challenges must be addressed in enabling high accuracy information extraction

under constraints on power consumption, computational resources, and latency. In this short course, we cover concepts and methods that help to address these challenges in the design and implementation of embedded SPML systems.

The short course is structured in two parts. The first part covers fundamentals of model-based design, which is an important foundation for developing efficient embedded SPML systems. In particular, we cover basic concepts of high level dataflow models of computation with emphasis on synchronous dataflow techniques. The second part of the course focuses on design issues for embedded SPML systems, and on emerging new

model-based design methods that are targeted to this growing application area.