I'm presenting on Strongly Typed Deep Learning with #haskell in a few weeks to the #StLouis machine learning and data science meetup.

I wish that #haskell's #tensorflow story was a bit better. I see a lot of potential for some really great capabilities to emerge from the combo of a language great for DSLs and with a lot of correctness guarantees, but the current solutions seem to require too much #categorytheory for a gentle intro to non-haskellers

Any suggestions for ways to make it easier?