Tom Breloff

Yes my first pass will focus on binary unums. Can you comment on potential pitfalls of this spec: ``` | exponent | fraction | signbit | ubit | ``` Questions:...

I re-read your comments and did some thinking, and I was wrong to get excited about the IntUnum concept. It would have to be a separate implementation, it couldn't use...

Thanks John! As I read the book there were some concepts that just "made sense" to me, and it's great to have a very easy to read book with enough...

Cool... Can you expand on how you are doing padding/alignment differently? If you have any other thoughts on implementation please let me know... It's not set in stone yet. Just...

Just so you're aware... OnlineAI is an experimental testbed... Not meant for public consumption. You should see LearnBase.jl, MLDataUtils, and the JuliaML org for the public interfaces we're working on....

Yes nothing stopping auto diff as part of the framework... Should be no more than a short method or 2 to implement. Let me know if you have a specific...

Thanks for the thoughts/links. When you say "3/7 types of layers", what exactly do you mean? My understanding (after a few readings of the paper) is that every LSTM component...

> but inputs, outputs, and bias units/connections do need special handling I'm not convinced yet. An input layer is a `GatedLayer` with connections to the (LSTM terminology) output layer, memory...

I've been mulling this over for a couple days. I have a good framework of types/structure without the forward/backward code, but I've been re-deriving the core math from a slightly...

I'm redesigning from my initial code to effectively represent unums similar to floats/ints in julia... using bitstypes: ``` # B is the base of the exponent, E is the number...