A toy model of Friston's active inference in Tensorflow
An attempt at implementing a toy model of a simple cognitive agent minimizing its informational free energy through variational inference (as described by Karl Friston’s Free Energy Principle). This model is highly inspired by McGregor et al., 2015.