Python version of the (matlab) MinimimalTransitionProbsModel.
https://github.com/florentmeyniel/MinimalTransitionProbsModel
This toolbox computes posterior inferences from a sequence of observations.
The inference can be:
- Bayes optimal solution assuming no change point
- with perfect memory
- with an exponential leak on observations count
- within a sliding window of observations
- Bayes optimal solution assuming change point (computed with Hidden Markov Model and numeric integration).
The Python version is more general:
- it is not restricted to the binary case: it can handle an arbitrary number of item
- it is not restricted to order-0 and order-1 transition probabilities: it can handle an arbitrary order
NB: due to numerical overflow, the inference will crash for higher order and large number of items.