0.3.0-alpha

• Refactor so loss can be an arbitrary function
• Fix bugs in and expand options for projetion
• prep-like CLI to prepare data for projection onto a trained model
• cellscore fraction file for score CLI
• Verbose option for load_txt
• Update options for validation cells & selection
• Version as an object attribute
• Handle change in scipy API
• new GENCODE files
• (feature request) options to specify a and c from the train CLI
• Documentation with ReadTheDocs

v0.2.4-alpha

Emergency patch preprocessing error for loom files. Also fixed an errant test. Not really enough to justify a new release but fixed a pretty irritating/embarrassing error.

v0.2.3-alpha: Merge pull request #5 from simslab/dev

• Emergency fix no split on dot error
• Max pairwise table + default max pairwise in score
• Note about ld.so error
• Fix max pairwise second greatest bug
• Some integration tests

v0.2.2-alpha

from dev pull request:

• partial test suite
• max pairwise method
• faster preprocessing of large text files
• refactor preprocessing and training control flow out of CLI
• move load and save methods outside of scHPF object
• message_function now checkstep_function in scHPF._fit

v0.2.1-alpha

• Slight speedup during inference (for Xphi calculation)
• Fix bug (occurred first in 0.2.0-alpha) genes in whitespace-delim input to prep that have no counts anywhere causes shape mismatches

v0.2.0-alpha

Numba implementation with scikit-learn-like API

v0.1-alpha

Tensorflow implementation