Introduction to Applied Machine Learning for Audio: Theory and Implementation

The contents of the course "Introduction to Applied Machine Learning for Audio: Theory and Implementation" are designed to convey concepts and understand classical machine learning methods with the goal of training models in a supervised or unsupervised manner, capable of classifying different types of audio, whether into single or multiple labels. Besides, we visited some public generative models using recurrent neural networks and LSTMs. The course is largely a compilation of information scattered across the internet, which has been translated into Spanish, with an emphasis on some demonstrations and mathematical tools inherent to the models, along with notebooks and examples oriented towards audio-related problems. The lecture is tailored for Sound Engineering students at the University of Chile.

Upon completing the course, the student will:

• Understand general ideas about the use and operation of different machine learning models.

• Understand different mathematical perspectives of these models, associating different mathematical tools as part of the different models.

• Recognize and use preprocessing strategies for audio handling.

• Apply these models to audio classification problems by using various strategies and comparing results using different statistical tools and data visualization techniques.

References

• Introduction to Probability for Data Science, Stanley H. Chan, 2021, Michigan Publishing. ISBN 978-1-60785-747-1 Bishop, Christopher M., and Nasser M. Nasrabadi. Pattern recognition and machine learning. Vol. 4. No. 4. New York: springer, 2006.

• Python Crash Course: a hands -on project-based introduction to programming. Second Edition. ISBN-13: 978-1593279288

• Deep Learning, Ian Good Fellow, Yoshua Bengio and Aaron Courville. MIT Press, 2016. ISBN, 0262035618, 9780262035613

• Neural Networks and Learning Machines, Haykin Simon, 2008. Third Edition ISBN 10: 0131471392 ISBN 13: 9780131471399.

• PhD Valerio Velardo: https://github.com/musikalkemist/AudioSignalProcessingForML

• Approaching (ALMOST) Any Machine Learning Problem, Abhishek Thakur, 2019.

• Deep Learning with Python, François Chollet, 2021, Second Edition, Manning Shelter Island.

• Python DataScience Handbook, Essential Tools for Working with Data. Jake VanderPlas O’Reilly, 2016, First Edition.

• PhD Professor Andrew NG: https://www.coursera.org/specializations/machine-learning-introduction

• PhD Keun WooChoi: https://github.com/keunwoochoi/dl4mir

• PhD Joshua Starmer: https://statquest.org

• PhD Marcos E. Valle (Desafio 1): Link

• Dive into Deep Learning

Reference Databases

1 - RAVDESS

2 - Covid-19

3 - Music Genre

4 - Free-Spoken-Digit-Dataset

5 - Karnika Kapoor Audio Generation

6 - The Independent Code