toy_data

Toy Experiments

Here we include the code for our analytical toy experiments. We use a toy mixture model and a toy HMM model for analyzing the properties of our McBits coders including:

• convergence: the convergence of net bitrate to the entropy as N increases
• cleanliness: the gap between net bitrate and ideal bitrate
• initial bit cost: the gap between total bitrate and net bitrate

You can play around with the toy examples in toy_example.py. We also include the notebooks for making the plots in our paper:

• initial_bits_plot_toy_mixture: the initial bit cost plot on toy mixture model
• cleanliness_convergence_plots_toy_mixture: the cleanliness anc convergence plots on toy mixture model
• cleanliness_convergence_plots_toy_hmm: the cleanliness anc convergence plots on toy HMM model