## Google Custom Search

Wir verwenden Google für unsere Suche. Mit Klick auf „Suche aktivieren“ aktivieren Sie das Suchfeld und akzeptieren die Nutzungsbedingungen.

## Coding for Private Reliable and Efficient Distributed Learning

Lecturer (assistant) 0000002680 Lecture 4 SWS Wintersemester 2023/24 English See TUMonline See TUMonline

### Objectives

At the end of this course, the students are able to: - Identify the challenges of modern distributed computing systems and tackle them using coding theory - Analyze and assess the privacy and the performance guarantees of coding theoretic methods applied to distributed computing - Manipulate the fundamental secret sharing tool and be able to apply it in several applications to private distributed systems - Understand the basics of machine learning algorithms - Implement through programming languages algorithms explained in a research paper, produce their own observations and analyze their findings - Prepare and present a poster (or a paper) within the standards of international conferences/workshop

### Description

This course covers the coding-theoretic tools behind speeding up distributed computing. In addition, data privacy being of crucial importance, this course includes a deep understanding of the main tool behind information-theoretic privacy: secret sharing. Several applications of secret sharing and coding theoretic tools in modern distributed systems and distributed machine learning are then explained together with the impact brought by applying those tools. More precisely, the course covers the following topics - Coding to ensure information-theoretic privacy through secret sharing. - Communication efficient secret sharing: a modification of secret sharing to better suit modern distributed systems - Basic probability theory such as ordered statistics used to analyze the performance of different techniques used in coding for distributed systems - Coding for private and fast distributed matrix multiplication, based on secret sharing schemes. Several state-of-the-art techniques will be explained in detail and their performance will be analyzed - Basics of machine learning algorithms: gradient descent, stochastic gradient descent, linear and logistic regression - Fast distributed gradient descent with and without coding theory

### Prerequisites

Basic knowledge of mathematics, e.g., linear algebra, calculus and probability theory Basic knowledge of a programming language, e.g., MATLAB or python The course "Channel Coding" is recommended

### Teaching and learning methods

Lectures: The lecturer uses slides and the blackboard (or an iPad) to explain the fundamental theoretical concepts. Lectures are designed to be interactive and to encourage the students to ask questions and initiate discussions with the lecturer and each other. Tutorials: The lecturer, or teaching assistants, solve concrete problems to simplify the understanding of the theoretical content and help the students better understand the material. Students’ presentations: Each group of three students nominates one of them to present a paper of their choice. The paper is recent research on topics discussed in class. (Groups of four students nominate two students, each presents a paper) Poster session: A workshop-style session in which each group of three students nominate one of them to present a poster. The others would go around and ask questions about other posters. This session is designed to introduce the students to environments of international scientific events. Throughout the semester, the students can ask for question-and-answer sessions and/or lab sessions in which the lecturers and teaching assistants come to answer the students' questions about the topics discussed in class. In lab sessions, the teaching assistants provide help to further their progress in the programming tasks.