Jesse Buesking

@TheKidCoder Ideally this wouldn't replace anything within Chronic but would be offered as an alternative implementation of the Numerizer logic. For example, we could have a gem that simply monkey...

@f0k Your version implements the `projection shortcut`, or version `(B)` in the paper (projecting using a 1x1 convolution when stride=2). How would you go about implementing the zero-padded version, version...

Also, doing a convolution with stride 2 results in much slower performance. Using `Conv2DCCLayer` should work around this [from what I've read](https://groups.google.com/forum/#!topic/pylearn-users/DDzJrE_lm4g), however it errors out with ``` ValueError: stride...

@ebenolson yep, `Conv2DDNNLayer` works well, thank you! @benanne I have yet to try it out, but thank you for pointing me in the right direction!

I agree, it does sound weird. Based on the code above, ``` def residual_block(layer, num_filters, filter_size=3, stride=1, num_layers=2): conv = layer if (num_filters != layer.output_shape[1]) or (stride != 1): layer...

Thank you! > I wonder why C? for some users that might seem unpreferable. I wrote a bit about the project [on my blog](http://jessebuesking.com/articles/programming/c/ruby/2017/02/13/omnomnum#benefits-of-the-new-approach), where I explain why I chose...

I agree, there's a decent bit of code broken out across quite a few files 😄 For one thing, C is much more verbose than ruby and requires more lines...

Some of the files were taken from the [Scipy](https://github.com/scipy/scipy) project and would fall under their licensing, and the TrueSkill algorithm itself is patented and trademarked by Microsoft so you would...