Xirt

Fixes the processing of linear peptides which is now possible again.

Minor bux fix release with doc improvements

- fix plotting issue when only one dimension is used for prediction
- add version parameter to main
- add readme as output
- remove pngs from output dir

Minor usabiility improvements and changes to the documentation.

- only load tensorflow when executing any work
- remove tensorflow-gpu dependencies
- remove graphviz and pydot dependencies

v.1.2.2
Minor Release update which improves the generation and documentation of the parameter files.

E.g. example parameter files are now available in the github repository and they can be generated by running:

>xirt -s learning_params.yaml
>xirt -p xirt_params.yaml

New Release with a couple of minor improvements:

- versioning is automated
- readme and docs
- example parrameters

*Feature**
- allows transfer-learning when the number of fractions is not equal to the training data

Major bug fixes/enhancements in

- ordinal regression encoding
- adds qc functionality to the main function
- enhances polynomial feature generation (only crosslinked peptide prediction errors are combined, as well as only peptide1/2 errors are combined) - for crosslink searches
- fixed heatmap plotting
- fixed 1-off error for softmax/sigmoid activation functions.

First official release.

Package is not yet fully representing all the options from the publication (work in progress).

This release of xiRT comes with the ability to:
- use any combiation of activiation functions / chromatography experimental setup
- use the convenient command line interface *ximl* to perform RT prediction for linear and crosslinked peptides
- encode crosslinked peptide input in a highly configurable Siamese network structure