[mlpack-git] [mlpack] Create a function to visualize the features learned by sparse autoencoder (#465)

[Ryan Curtin]

Sure, it would make the class more complicated, that is true.  Personally I tend to prefer allowing more general code, but providing defaults where possible.  I think that allowing the user to set `blockWidth` and `blockHeight` is really important, because if the user is making an autoencoder for images, the images can be of arbitrary size.  An image with 1024 pixels could be 32x32, or it could be 64x16; we can't know what the user wants there, so in that particular case I don't think it even makes sense to provide a default parameter.  If the default parameter is wrong (i.e. if the default selects 32x32 but the images are actually 64x16), then the user gets back a visualization matrix that looks wrong and isn't very useful for visualization.

Does my idea make sense here, or maybe am I misunderstanding something?

---
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/465#issuecomment-158938695
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20151123/f8888341/attachment.html>