Special Topics: Machine Learning and Audio: a challenge
GeneralThis course will be taught in English. (German is possible, if there are only native speakers)
PrerequisitesThis course requires (extensive) prior experience with machine learning and familiarity with deep learning frameworks such as Theano, Tensorflow and PyTorch. Programming experience in Python is therefore also a non-negotiable requirement. Furthermore, code will be versioned and shared via GitLab, hence a familiarity with Git and GitLab is a big plus.
Course OutlineDepending on the number of participants, we will form one or more teams. We will choose to either tackle a challenge related to audio and music on a platform such as Kaggle, or work on one or more MIREX tasks that we collectively deem interesting. Each group will implement a complete data processing pipeline from raw data and annotations to final predictions and including evaluation. Ideally this pipeline will include methods to facilitate the tuning of hyper parameters and model selection. We will of course assist you during this whole process with suggestions and helpful tips on methods and general approaches to such projects. We will provide datasets if we can, or help finding and distributing them, and will also be able to provide a limited amount of (shared!) computing resources.
ModeWe will have weekly meetings where you will give a small report (5 slides at most) on your progress so far. Presenters will be selected from each group in a round-robin fashion. Sporadically, we may also review code, and ask individual group members clarification questions. Each group will need to produce a final, written report, detailing their approaches and lessons learnt. The challenges will be done in teams and although the final outcome is not that important, winning one or all of them would be a big plus (for you, your team, and the university). You will be graded individually. The grade is based on attendance, presentations, questions answered, and contributions to code and final report.