Seminar: Parallel computing in machine learning
In machine learning, tasks like parametric search or cross-validation are time intensive. In this seminar, we will explore how to use multithreading, multiprocessing, and compute clusters to reduce the execution time of machine learning frameworks. Additionally, we will cover some python basics and patterns to simplify parallel framework development.
The seminar has a heavy practical component to practice and become familiar with the challenges of parallel data processing and machine learning. Therefore, we invite participants to bring their own dataset for analysis, otherwise, we will provide a dataset for exploration.
- Gain an overview of general parallel processing tools and techniques.
- Understand concepts of data processing, job distribution in machine learning frameworks.
- Analyse machine learning frameworks in terms of data storage and processing.
- Apply the parallel job approach to bigData problems.
- Implement a distributed job handling solution for bigData and machine learning.
|Presence time||Lecture time: 2 SWS, exercises: 3 SWS|
|Useful knowledge||Python and data analytics|
|Starting date||Winter semester 18/19|
Up-to-date literature recommendations are provided during the lectures.
- Final project presentation, demonstrator and final report.