ConstrainedHiddenMarkovModel

This novel model was implemented as part of my Master's thesis to generate musical sequences styled after Bach chorales (also see https://github.com/po-gl/BachPipeline). For example, the sequence below is constrained to be the beginning and end of the first five measures of Bach's "Wer nur den lieben Gott läßt walten". Constraints are colored green.

The constrained hidden Markov processes is an extension of work done by Pachet et al. in their paper, "Finite-Length Markov Processes with Constraints." The model generates sequences and can apply user-defined constraints to the sequences. Sequences can be generated in any number of domains, such as natural language or music generation.

Above is a visual representation of a trained constrained hidden Markov model.

Usage

Ensure that cargo is installed then run using the following command inside the project directory.

cargo build --release
cargo run -- -n 10 -c config.yaml

Available options are listed below:

USAGE:
    constrained_hmm [OPTIONS]

OPTIONS:
    -c, --config <config_file>     YAML config file path
    -f, --file <training_file>     Training file path
    -h, --help                     Print help information
    -m, --order <markov_order>     Markov order
    -n, --sequences <sequences>    The number of sequences to generate
    -o, --out <output_file>        Output file to write sequences to

Constraints are specified by the YAML config file. See config.yaml for an example.