Blahut Arimoto Algorithm

What we are trying to achieve with the implementation of the Blahut-Arimoto Algorithm (bottleneck algorithm) is to find the model that compresses the information the most, while decreasing the distortion, i.e keeping relevant our model, and preserving as much information as possible. By changing the value of the parameter \beta we can actually explore the trade-off between compression and the preserved meaningful information.