Made an mlp baseline that actually scores surprisingly well in ndcg on the large test set. It's competitive with our other top models.

[erikhuck]

closes #184

[codecov-io]

Codecov Report

Merging #185 into develop will increase coverage by 2.1%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           develop     #185     +/-   ##
==========================================
+ Coverage    54.28%   56.38%   +2.1%     
==========================================
  Files           35       35             
  Lines         2487     2497     +10     
==========================================
+ Hits          1350     1408     +58     
+ Misses        1137     1089     -48

Impacted Files	Coverage Δ
dna/models/__init__.py	100% <ø> (ø)	⬆️
test/test_models.py	94% <100%> (+0.38%)	⬆️
dna/models/baselines.py	35.71% <100%> (+2.57%)	⬆️
dna/models/base_models.py	63.66% <0%> (+17.26%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d2abe98...c28be61. Read the comment docs.

[bjschoenfeld]

You should not use the final test set until we have had a chance to tune and finalize the model.

Please add some tests to show that the new model is working.

[erikhuck]

@bjschoenfeld I didn't use the final test set. I was referring to the development test set. The one that gets scored every time we run.