The Genetic Algorithm, Neural Networks, Feature-Extraction and Use
Still playing catch-up from the hurricane, but we're almost there.
We covered three broad areas in this class: the Genetic Algorithm
(meaning a particular approach to algorithmically-solving problems,
the underlying concept drawn from basic genetics/evolution theory),
neural networks, and feature-extraction and subsequent use. All
three of these approaches are a little different from the past techniques
we have been covering in that they aren't data-generating algorithms
per se. Instead they are techniques that may be employed in
the construction of music in various ways, decision-making,
categorization, etc.
Links
Class Downloads
- week11-classpatches.zip
-- this archive contains three sub-directories:
- genetic -- this has the Xwindows code and MacOSX exectuables
for the 'sine-wave convergence' GA I showed in class. It also includes
the two example soundfiles (mutate.wav and nomutate.wav)
generated by the software. It should be compile-able on Windows with
the appropriate Xwindows package installed.
- nn-timbre -- this has the Xwindows code and MacOSX exectuables
for the simple neural-network timbre classifier I showed in class.
- iglesia-net -- Daniel Iglesia's Max/MSP version of the
simple neural-network timbre classifier.
- meap.zip
-- Bryan's MEAPsoft demo examples.
- mitchell-bach.mp3
-- Bryan's snazzy 'bach-played-by-Roscoe-Mitchell' soundfile.