Evolutionary Hidden Markov models (HMM)
For population training, run:
python main.py -c config.cfg -m song.mid
For new midi file generation, run:
python generate_midi.py -b model.pickle -m song.mid -s new_song
Config file
- populationSize - size of population
- generationCount - number of generations
- crossoverFraction - witch fraction of population combine
- randomSeed - random seed
- learningRate - learning rate
- nHiddenStates - number of hidden states in HMM
- nObservableStates - total amount of observations
- useFullSequence - if true will use nObservableStates, else will use unique values in observed sequence