Introduction to Machine Learning
Introduction to Machine Learning
Abstract:
The course offers a gentle, hands-on introduction to the core principles and techniques of machine learning. The participants will be presented with the basic supervised machine learning models (linear models, trees and some instance-based techniques), along with the principles of their application (training, testing, cross-validation, regularization, model selection). Students will implement and apply most of the discussed algorithms using the interactive environment IPython. This will be followed by a discussion of the general principles underlying machine learning, such as statistical estimation methods and optimization techniques. Some extent of basic familiarity with programming and Python is expected from the participants. Previous knowledge of probability theory and linear algebra would be helpful and could ensure better understanding of the material.
Associated projects:
Students who would take extra interest in this course are offered a number of projects to pursue during the “practical” part of the Summer School (in addition to perhaps more topic-specific data-oriented projects given by other instructors). The following is a tentative list of topics for inspiration. For any of the topics the exact details of the project will be clarified in collaboration with the students that decide to take it on. All projects imply either a practical implementation in the form of usable software or a written report on the performed steps, experiments and results.
- Machine learning 101: Handwritten digit classification study.
- Can a computer guess your language? Adaptive clustering of spoken text.
- Priority inbox: Predict the importance of messages in your mailbox.
- Computer-generated music or text: Train a generative model to compose songs or lyrics.
- Summarize the news: Create an informative visualization of news articles (using SOM, wordclouds, decision trees, or the like).
- Recommended reading: Implement a simple recommendation system for the posts of a forum.
- Contribute to the community: Implement an algorithm in Spark MLLib that is not yet implemented there.
- Show off what you learned: Implement a javascript demo/learning tool for an algorithm of interest.
- etc
About lecturer:
Mr. Konstantin Tretyakov,
Researcher, University of Tartu, Estonia.