Added
- made into separate project
- set up repository with structure from "Good Enough Practices for Scientific Computing"
<https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1005510sec009>
- add ability to run learning curve to test whether train / test accuracy are converging
+ i.e. would we benefit from more training data
- can compute accuracy per epoch per set size on training set, saves to .txt file
Fixed
- `__main__` checks whether config file exists and throws a human-readable error if it doesn't,
as opposed to failing weirdly because `ConfigParser` doesn't tell you when the `.ini` file does not exist