News for hackers. Stuff that matters.
Re:Every Model Learned by Gradient Descent Is Approximately ... by scottlocklin on Saturday December 05, 2020 @09:40AM (#25314834)
Looking at you, deep learning.
Re:Every Model Learned by Gradient Descent Is Approximately ... by wongarsu on Sunday December 06, 2020 @10:34AM (#25324119)
A linear SVM can in turn be expressed as a very shallow neutral network. The main difference is that with SVMs you put all your effort into transforming inputs for the model (e.g. all the popular kernels) while with neural networks usually most of the effort goes into clever model architectures.
Re:Every Model Learned by Gradient Descent Is Approximately ... by api on Sunday December 06, 2020 @10:27AM (#25324055)
There is probably a ton of isomorphism between different models. It may come down to what is easiest to understand and fastest to implement in code.
Re:Every Model Learned by Gradient Descent Is Approximately ... by segfaultbuserr on Sunday December 06, 2020 @10:34AM (#25324118)
See also:

A visual proof that neural networks can approximate any function

https://news.ycombinator.com/item?id=19708620

Re:Every Model Learned by Gradient Descent Is Approximately ... by api on Sunday December 06, 2020 @11:51AM (#25324657)
So a "neural network" is actually a type of parameterized mathematical function that can be fit to any curve including higher dimensional surfaces, etc.?
Re:Every Model Learned by Gradient Descent Is Approximately ... by xksteven on Sunday December 06, 2020 @12:42PM (#25325087)
Re:Every Model Learned by Gradient Descent Is Approximately ... by laingc on Sunday December 06, 2020 @12:41PM (#25325079)
Yes.