Back to TUM-PREP Projects overview

Project Overview

Project Code: ED 29

Project name:

Coding for Distributed Machine Learning

TUM Department:

ED – Engineering Sciences

TUM Chair / Institute:

Coding and Cryptography

Research area:

Coded Computing

Student background:

Computer EngineeringComputer ScienceComputer Science/ InformaticsElectrical EngineeringMathematics

Further disciplines:

Participation also possible online only:

Planned project location:

Technical University of Munich
Theresienstrasse 90
80333 München

Project Supervisor - Contact Details


Title:

Given name:

Christoph

Family name:

Hofmeister

E-mail:

christoph.hofmeister@tum.de

Phone:

+49 (89) 289 - 29057

Additional Project Supervisor - Contact Details


Title:

Prof. Dr.-Ing.

Given name:

Antonia

Family name:

Wachter-Zeh

E-mail:

antonia.wachter-zeh@tum.de

Phone:

+49 (89) 289 - 23495

Additional Project Supervisor - Contact Details


Title:

Given name:

Maximilian

Family name:

Egger

E-mail:

maximilian.egger@tum.de

Phone:

Project Description


Project description:

A rapidly growing body of research deals with applying coding and information theoretic ideas to distributed computing and especially machine learning [1]. The aim is to tackle the following three fundamental challenges:
i) Privacy: When training a machine learning model, how can the privacy of sensitive training data be preserved?
ii) Stragglers: The computation performance and data transmission speeds of the nodes in the system is typically heterogeneous and time varying. How can computations be distributed efficiently in a way such that a few slow nodes do not degrade the performance of the whole system.
iii) Security: Some of the nodes in the system might be malicious and send incorrect responses. How can such nodes be detected and prevented from affecting the overall computation result?

The student's task is to study and/or implement a scheme in the field of coding for distributed machine learning.

[1] Li, S., and Avestimehr, S. 2020. Coded Computing: Mitigating Fundamental Bottlenecks in Large-Scale Distributed Computing and Machine Learning. Found. Trends Commun. Inf. Theory, 17(1), p.1–148.

Working hours per week planned:

35

Prerequisites


Required study level minimum (at time of TUM PREP project start):

3 years of bachelor studies completed

Subject related:

Linear Algebra, Probability Theory

Algebra, Coding Theory and Information Theory as well as programming experience would be helpful but are not required.

Other:

  • Keine Stichwörter