Speedy Composer

Welcome to

Speedy Composer
(Artificial Neural Network Melody Composer)

Contents:

Short Summary

A non-technical summary of the project.

Main Documentation.

Introduction.
Background.
The network architecture and output.
Representing the input for the networks.
The training set.
The training algorithm.
Comparing the networks to each other.
Composing new melodies.
Conclusions.
Future research and projects.
Acknowledgements.

Appendix 1 - Examples of melodies composed by Speedy Composer:

Compositions without a given pace (using the note_begin network).
Compositions according to a given pace (without using the note_begin network).

Appendix 2 - The training set:

Statistics of the training set.

Appendix 3 - List of the source files used in Speedy Composer:

MATLAB m-files.
C++ files.
Visual Basic files.

Appendix 4 - Examples of the networks output.
Appendix 5 - MATLAB Graphs:

A graph of the pitch_probability_array.
Graphs of the error function when training with TRAINRP.

Appendix 6 - The User Interface.
Appendix 7 - Statistics of the training set and networks performance for June 1999.
References.

Recommended books about computer generated music.
Recommended books about Artificial Neural Networks.
Recommended websites about music.
Recommended websites about Artificial Neural Networks in general.
Recommended software for training or simulating Artificial Neural Networks.
More articles about Artificial Neural Networks and music composition.

Best compositions of Speedy Composer:

Files to download:

(The composing program will be available in the future).

Compositions.zip (22KB) - Examples of melodies composed by Speedy Composer (in MIDI format).

Speedy Composer websites:

Please tell me what you think about the composed melodies - I'm interested in your opinion. you can contact me at uri@speedy.net .
Uri Even-Chen.