v0.7.1

Minor bug fixes.

New parameters for BW.fit:

• verbose, bool (default is True): Print or not a small recap at the end of the learning.
• return_data, bool (default is False): If set to True, a dictionary containing the following values will be returned alongside the hypothesis once the learning is done. 'learning_rounds', 'learning_time', 'training_set_loglikelihood'.
• stormpy_output, bool (default is True): If set to True the output model will be a Stormpy sparse model. (not available for GoHMMs and MGoHMMS).

BW.fit method can take a list or a numpy ndarray as a training set. It will be internally converted to a jajapy.Set.

v0.7

• Model checking for shortest path properties on MDPs
• BW_HMM improved
• Small summary after learning
• Can set max number of iterations while using BW
• Documentation enhanced, addition of the "formal" page

v0.6 Model converter to stormpy

A new function jajapy.modeltoStorm converts any jajapy.HMM, jajapy.MC, jajapy.MDP and jajapy.MC to the corresponding model in stormpy. Notice that these models are trace-equivalent, but not similar: stormpy models have more states than jajapy models, since labels are on the states in stormpy while on the transitions in jajapy.

New examples have been added.

v0.5.1

Small correction for setup.py.
Now pip install works properly

Restructuring of models

All the models are now explicitly described with numpy array. It's a little bit worst from the memory perspective but it's a lot better in terms of running time.

The documentation have been updated and improved.

v0.4

A new way to create model states.
Now, to create a MC state with the two following transitions:

• To state 1, observing 'a', with probability 0.6
• To state 1, observing 'b', with probability 0.4

it's
MC_state([(1,'a',0.6),(1,'b',0.4)],0)
instead of
MC_state([0.6,0.4],[1,1],['a','b'],0)
0 being this state ID

v0.3

Units tests ✅
More examples in the documentation

v0.2

Documentation is here!

v0.1.1

First release.