Navigation

Approximate Computing

Lecturers:

Prof. Dr. O. Keszöcze

Prof. Dr.-Ing. J. Teich

 

Module description:

Approximate Computing and
Exercises to Approximate Computing

Lecture, location and time:

Thursday, 10:15 – 11:45 h. Room 01.255-128 Registration on StudOn

First lecture will be April 25th

Exercises, location and time:

Tuesday, 10:15 – 11:45 h. Room 02.133-128 (J. Echavarria)

Course (slides, exercises, other files):

All important documents can be found in StudOn

 

Introduction:

Approximate Computing denotes a quite young research area that exploits the fact and capability of many applications and systems to tolerate imprecision and/or inexactness of computed results. Prominent areas of applications and novel techniques of computing approximate rather than exact results have brought up new implementations either at hardware and/or software levels for important emergent workloads such as searching, mining, image processing, and data retrieval.

Although hardware technology is improving at a fast pace, energy and power are becoming more and more important constraints apart from exactly computing results in an acceptable amount of time. The main goals of approximate computing techniques are therefore to exploit the possible trade-off between power/energy consumption, accuracy, performance, and/or cost, e.g., utilized hardware resources.

Course purpose:

The purpose of the course approximate computing is to instruct students about the main ideas and concepts of approximate computing. This includes analyzing the trade-off between energy consumption, accuracy, run-time and hardware costs, concrete approximating techniques (e.g. approximate hardware synthesis, approximating algorithms) as well as theoretical background (determining the computational error and its complexity).

Course content:

The following topics will be covered:

  1. Introduction to Approximate Computing
  2. Challenges
  3. Approximate Arithmetics
  4. Approximate Algorithms
  5. Applications

Approximate computing is a lecture on a 4 SWS (4 hours/week) basis. Lecture and Exercises will give 5 ECTS.

Resources:

Approximate Computing: An Emerging Paradigm For Energy-Efficient Design

Approximate Computing: A Survey

Useful links: