Sam Lendle

I'm also on Arch. I'm getting a different error, but I'll report it here instead of opening a new issue. ``` ==> installing Node.js dependencies > [email protected] install /tmp/luarocks_gfx.js-scm-0-1584/gfx.js/node_modules/pty.js >...

This is not the described behavior in the [docs](http://numericextensionsjl.readthedocs.org/en/latest/reduction.html#basic-reduction-functions). Also, `mean!` function (which you can access by `NumericExtensions.mean!` because the package is not importing Base.mean!, for those following along at...

The link, inverse link, and mueta functions for ProbitLink() are correct I think so I am not sure where the trouble is coming from.

The k-medoids code was rewritten by @cyocum in #22.

I just looked at the implementation and docs. C is actually an [n x n matrix](https://github.com/JuliaStats/Clustering.jl/blob/f83803d10e78e549dfb3938eda78524274403287/src/kmedoids.jl#L30-L31) (not k x n). The docs say "C – The cost matrix, where C[i,j]...

@JohnHBrock I think you need to be using torchdata's DataLoader2, not DataLoader.

What about using different names like `nfo` and `wrn` (or something better)?

Wouldn't the gradient require knowledge of some link function (maybe determined by `RegressionProblem`) as well? In `tloss_and_gradient` you're calculating the gradient as `X * dv`, but in GLMs for example,...

After reading again, it looks like the argument `f` in `value(::Loss, f, y)` is expected to be linear in the coefficients. Is that correct? In that case, the gradient calculation...

@lindahua By nonlinear regression I meant solving the optimization problem ``` latex \sum_i f(h(\theta, x); y) + r(\theta) ``` for some known, nonlinear `h` as in nonlinear least squares.