Very cool. As someone who wants to begin learning how to use ML and who enjoys clojure as a hobby language...this is great. Gives me an excuse to use Clojure more :) I hope you decide to continue working on it!
I'll try! I've got a full time management job so this was a fun way to code again. The library is pretty small and I think easy to follow if you want to learn
This made me chuckle. If anybody remembers the dim and distant past of various singularitarian mailing lists on the internet, you might remember Eliezer Yudkowsky's theoretical 'annotative' language Flare[1]. It was one of the early steps on the (what was then quite straight in some people's minds) line to AGI.
As for this, possibly worthwhile for the authors to look into Dragan Djuric's work, he's done some really fast matrix stuff on Clojure[2].
What is the optimization algorithm for these if not backpropagation? This page does mention the use of hyperparameters for optimization and compares itself to an adaptive stochastic gradient descent method, which makes me think that it is using backpropagation.
Thinks for your great lib in both great Clojure and ML. Do you know whether this Clojure lib can do Dynamic net: https://github.com/thinktopic/cortex? you didn't compare it in your blog.
Yes this does dynamic nets, the model is the same as PyTorch, you build a graph for each input(s). I think Cortex is in the Keras-style of model (only layer abstractions and not "dynamic" in the sense we talk about neural nets) and isn't dynamic, but I could be missing something. The examples seem to all be in that mold: https://github.com/thinktopic/cortex/blob/master/examples/ca...