Skip to main content

Artificial Models for Music Creativity: Part 1

Fach laut Studienplan
CoPeCo, E-1-Jko-MM, Tec-W1-MMK, Tec-W2-MMK, W-bv, W-frei, Wiss-1-MMK
Lehrende
Semester
Wintersemester 2023/2024
Termin
From 10 to 17 the following days: Sat. 14.10 - MFS Sat. 4.11 - MFS Sun. 3.12 - MFS Sat. 6.1 - MFS Sat. 3.2 - to be defined
Raum
Multifunktionsstudio (ehemals Schauspielstudio 3)
Dauer
1.5 Semesterwochenstunden
Beschreibung

This course offers a hands-on experience and provides a foundational theoretical framework for integrating machine learning models into artistic practices, with a specific emphasis on music scenarios. The course spans two semesters, each comprising assignments and a laboratory component. "Artificial Models for Music Creativity" is scheduled as a block seminar once a month in each semester.

During this course, we will delve into the exciting world of machine learning and its application to artistic endeavours, particularly in the realm of music. Our focus will be on three major types of models extensively used in Magenta: Recurrent Neural Networks (RNN), Variational Autoencoders (VAE), and Generative Adversarial Networks (GAN).

Semester 1: Introduction to Symbolic Data Representation
In the initial part of the course, drawing inspiration from artistic examples, we will explore the fundamental concepts of working with symbolic data representations, such as MIDI. This essential knowledge will equip you with the tools necessary to unleash the creative potential of artificial intelligence in the context of music. The seminar seamlessly integrates theory and practice, guiding you through a diverse range of topics, including:

Introduction to Python
Setting up Virtual Environments
Jupyter Notebooks
Essential Libraries for Machine Learning
Supervised Machine Learning: Regression and Classification
Introduction to Neural Networks and Feedforward Neural Network
Exploratory Data Analysis and Preprocessing Techniques
Train/Test Data Split
Cross-validation Methods
Bias/Variance Trade-off
Understanding Model Statistics and Visualization Techniques

Literatur

The reference literature will be provided by the instructor during the first lesson.

Credits
4 Creditpoints
Bemerkung

The course does not require prior knowledge of Python; however, interested students are encouraged to contact the instructor before the start of the class at alessandro.anatrini[at]hfmt-hamburg.de
The course is offered in person in the English language.

Module
CoPeCo, Multimediale Komposition (Jazzkomposition Master), Technisches Wahlmodul 1 Multimediale Komposition Master, Technisches Wahlmodul 2 Multimediale Komposition Master, Berufsvorbereitendes Wahlmodul Master, Wahlmodul freie Wahl (alle Studiengänge), Wissenschaftliches Modul Multimediale Komposition Master