v0.2.3

Changes:

• Added citation comments (#19)
• Put the RNG (self.rng) in HSMM class constructor (#17). This fixed a bug wherein sample() with specified random_state has the same output every time it is run.

This is the final release before I archive this repository for good. The above fixes are just for making the code easier to read for reference. Hope this project helped you in any way, but I have to move on and leave this in my mind.

poypoyan

v0.2.2

Changes:

• Support for parallelism with Joblib: All methods now return something. For fit(), it returns self, the fitted model.
• There is a new attribute called "name". This is used in displayed messages for monitoring. This is helpful when models are run in parallel.

v0.2.1

Changes:

• Left-censoring implemented. WARNING: experimental.
• Made the repo more "venv-friendly" for testing.
• Parameter rnd_state changed to random_state.
• PEP8 blank lines.
• Renamed variables related to log-likelihood.

I deleted v0.2.0 because of the above renaming of variables.

I may now leave this project for now.

v0.1.3

Changes:

• Removed all usage of ndarray.resize, and do pre-allocation instead. This is good practice.
• In iter_from_X_length() from hsmm_utils.py, raise error if end[-1] != n_samples. This is more general than before, which is just end[-1] > n_samples.
• Removed stray spaces.

v0.1.2

Honestly, I just want to release something. Further testing is needed.

There is no v0.1.0 because I messed up PyPI for that version. I deleted v0.1.1 because its source distribution is broken.