Аспирантура
2024/2025
Инструменты разработки для исследовательской работы в области глубинного обучения
Статус:
Курс по выбору
Когда читается:
1-й курс, 1 семестр
Охват аудитории:
для своего кампуса
Язык:
английский
Course Syllabus
Abstract
One of the major advantages of good code is that one can effectively use the models and run a ton of experiments with minimal efforts. In industry there exist a large set of common practices and tools to achieve an immense flexibility of the code and optimize the team efforts. Some of the tools can be readily applied in research as well. We suggest to consider a broader view on the ML development tools and study them in the context of several problems arising in pictures and sound applications. We will consider the code structure and planning, the tools for config management and result storage and processing.
Learning Objectives
- Tools of architecture design
- Solutions for configuration management
- Solutions for log management
- Solutions for documentation, git and deployment
Expected Learning Outcomes
- Code architecture design
- Project planning
- Work in collaboration
- Pipeline design and leveraging with prototyping
Course Contents
- Intro: experiment planning, python language hacks case study. From math to solution.
- Pipelines vs Prototyping: advantages and disadvantages, pipeline planning. Configuration management.
- Pytorch meets pipelines. How to construct the simplest DL pipeline. Zooming into torch.
- Git and collaboration case study.
- Auto-formed documentation, GIT pre-commit and auto code formatting.
- Workshop session: student presentations or invited talk
- Workshop session: student presentations
Assessment Elements
- Home Assignment 1: Code RefactoringAvailable starting from day 1.
- Home Assignment 1: Case studyAvailable starting from day 1. In this assignment the student has to present one of the code repositories of the scientific paper he uses extensively. In the presentation (up to 30-40min) it is required to - State the research goal the code implements - Make a review of repository structure: what are the usage scenarios, modules and how they are supposed to interact - What is good in the implementation and what can be suggested to improve the processes or the UX The presentations are scheduled for weeks 6-9
Interim Assessment
- 2024/2025 1st semester0.6 * Home Assignment 1: Case study + 0.4 * Home Assignment 1: Code Refactoring
Bibliography
Recommended Core Bibliography
- Rogers, S., & Girolami, M. (2016). A First Course in Machine Learning (Vol. 2nd ed). Milton: Chapman and Hall/CRC. Retrieved from http://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=edsebk&AN=1399490
Recommended Additional Bibliography
- A Tutorial on Machine Learning and Data Science Tools with Python. (2017). Retrieved from http://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=edsbas&AN=edsbas.E5F82B62