News for hackers. Stuff that matters.
Re:Every Model Learned by Gradient Descent Is Approximately ... by bicepjai on Sunday December 06, 2020 @12:22PM (#25324923)
Watch your stock NVIDIA :)
Re:Every Model Learned by Gradient Descent Is Approximately ... by rlili on Sunday December 06, 2020 @12:28PM (#25324968)
Why, aren't GPUs pretty useful in training SVMs as well?
Re:Every Model Learned by Gradient Descent Is Approximately ... by xksteven on Sunday December 06, 2020 @12:40PM (#25325071)
SVMs are solved via convex optimization methods which have taken more time to get on the GPU train.

On the other hand there are GPU accelerated SVM training such as: https://github.com/Xtra-Computing/thundersvm

A github or Google search will reveal other GPU accelerated SVM training.

Re:Every Model Learned by Gradient Descent Is Approximately ... by knuthsat on Sunday December 06, 2020 @12:53PM (#25325163)
SVMs can also be trained using gradient descent.
Re:Every Model Learned by Gradient Descent Is Approximately ... by abeppu on Sunday December 06, 2020 @12:39PM (#25325067)
Well, and though the author shows an approximate equivalence, which helps us understand a class of models better, it's not obvious that it's preferable to use SVMs of the type described. In particular, it seems like often it would be preferable to deal with model weights (even if they are mathematically a "superposition" of datapoints) than to ship around and revisit the whole dataset.