Statistical machine learning with applications in music
| Thesis title in Czech: | Statistické strojové učení s aplikacemi v hudbě |
|---|---|
| Thesis title in English: | Statistical machine learning with applications in music |
| Key words: | strojové učení, tensor flow, hudební skladba, neuronové sítě s LSTM, hodnocení hudby |
| English key words: | machine learning, tensor flow, music composition, neural networks with LSTM, evaluation of music |
| Academic year of topic announcement: | 2018/2019 |
| Thesis type: | diploma thesis |
| Thesis language: | angličtina |
| Department: | Department of Probability and Mathematical Statistics (32-KPMS) |
| Supervisor: | doc. RNDr. Jan Večeř, Ph.D. |
| Author: | hidden - assigned and confirmed by the Study Dept. |
| Date of registration: | 23.10.2018 |
| Date of assignment: | 23.10.2018 |
| Confirmed by Study dept. on: | 17.06.2019 |
| Date and time of defence: | 04.02.2020 08:00 |
| Date of electronic submission: | 06.01.2020 |
| Date of submission of printed version: | 06.01.2020 |
| Date of proceeded defence: | 04.02.2020 |
| Opponents: | doc. RNDr. Zdeněk Hlávka, Ph.D. |
| Guidelines |
| This work aims to review the current state of the art in statistical machine learning and apply it to music composition. The thesis should first briefly describe the general methods of statistical machine learning. Second, thesis should survey the existing methods applied in machine generated music. The current mainstream methods of statistical machine learning are widely implemented in Python, so the author of the thesis should get familiar with this programming language and the respective machine learning libraries, namely Scikit-Learn and Tensor Flow. In addition, Google Brain Team has recently released Magenta, a research project working with Tensor Flow aimed at machine learning applications for arts. Lastly, the work should train the computer to produce its own compositions based on a training data. The training music data are expected to be input as MIDI files from a selected set of existing music compositions. More specifically, one may focus on a small number of music composers in order to train the computer system to a more narrow data set, which can arguably lead to more pleasing outcomes. A possible set of training data can come from The Beatles songs or from a selected set of guitar solos, which are all well documented.
This thesis requires specific skills from the author, namely he/she should be both capable of working with data, but also should have a minimal understanding of music. |
| References |
| Efron, Hastie: Computer Age Statistical Inference, Cambridge University Press, 2016
Muller, Guido: Introduction to Machine Learning with Python, O'Reilly, 2016 Geron: Hands-on Machine Learning with Scikit-Learn and Tensor Flow, O'Reilly, 2017 The Beatles: Complete Scores, Hal Leonard, 1993 |
| Preliminary scope of work |
| Vytvořte hudbu se statistickým strojovým učením. |
| Preliminary scope of work in English |
| Create music with statistical machine learning. |