In order to learn more about the tech underlying DNN, I've been tinkering with a toy, very statically-typed, TensorFlow-like system in Scala. It is nowhere near ready for useful work and likely never will be, but lemme know if this piques your interest:https://github.com/avibryant/nda
-
-
Replying to @avibryant
Fun project! I saw another "typed interface to <deep-learning-tool>" project recently but predictably I can't find it now
1 reply 0 retweets 0 likes -
Replying to @manyangled
To be clear this (for better or worse, probably worse) doesn't wrap TensorFlow or anything else, it's Scala all the way down.
1 reply 0 retweets 0 likes -
Replying to @avibryant
Yes! I just figured you might find it interesting, what with all the typing and TF
... If I could find it
1 reply 0 retweets 1 like -
Replying to @manyangled @avibryant
It was the Haskell one that Jose Alonso tweeted, right?
2 replies 0 retweets 0 likes -
Replying to @willb @avibryant
Erik Erlandson Retweeted José A. Alonso
Erik Erlandson added,
1 reply 0 retweets 1 like -
Replying to @manyangled @avibryant
I’m just barely too slow! (Types and NNs is a conversation we need to have soon.)
2 replies 0 retweets 2 likes -
Replying to @willb @avibryant
If the input features are typed, they have to be the same type or dot-products won't compile! Either that or weights require differing types to compensate...
1 reply 0 retweets 0 likes
Aha yes! From what I can tell scanning this examples, this is going for similar properties to what I have (just in a haskell accent instead of scala).
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.