deep_music

This example tests the deep predictive learning model on predicting longer sequences with some structure that is not fully regular, by trying to predict the next note in a musical track. This requires representing longer-term sequences.

music

Initially targeting simple one-note-at-a-time songs ("melodies").

• bach_goldberg.mid: downloaded from http://www.jsbach.net/midi/midi_goldbergvariations.html -- v01

Performance

The model with 2 hidden + CT layers can learn 30 note steps to a high level of predictive accuracy -- near perfect for a spiking net. When trained on the whole song, it does not do as well, but improvisation testing reveals that it has absorbed a significant amount of song structure.

Mac playing the actual music from network

• open Apple MIDI Setup
• do Window / Show MIDI Studio
• double click on Default IAC Driver
• click Device is online -- the icon should now become undimmed
• open Garage Band (or Logic Pro if you have it)
• uncomment the Play setting in deep_music.go ConfigEnv method
• it should just work..