Using Neural Networks With Regression

http://deeplearning4j.org/linear-regression.html

by vonnik • 11 years ago

31 10 11 years ago

agibsonccc3• 11 years ago

Hey all,

Didn't expect to get on the front page. Just want to make sure we clarify what the intent was here.

A lot of people on our user group have been asking us about predicting continuous values in neural networks. This page rose out of the need of explaining that a neural network itself is a universal approximator that solves for different objectives/loss functions.

Two instantiations of that problem are regression and classification only changing w.r.t the activation function and objective function.

Happy to clarify/update anything.

Thanks!

haddr• 11 years ago

Side comment: great to see some java library for deep learning... I think so far it's mostly python and some C...

agibsonccc3• 11 years ago

Thanks! We'll be working on python bindings as well. We're finishing polishing the lib now.

Happy to help people on gitter as we finish this out:

https://gitter.im/deeplearning4j/deeplearning4j

ratsimihah• 11 years ago

Don't forget lua :)

pigscantfly• 11 years ago

Caffe also supports Matlab and C++, I believe.

microtonal2• 11 years ago

Caffe is in C++ and supports Matlab and Python. It's quite easy to bind, e.g. I have a Go binding that I use for 'neutral net surgery'.

hellrich• 11 years ago

With no ill intent: I'm a bit skeptical about software quality / test coverage after finding a bug in the linear algebra implementation - I would not use it for production right now.

Using Neural Networks With Regression

Related Stories