Unsupervised Feature Learning from Images on GPU

Hi,

I am trying to learn hierarchies of features on Images in an unsupervised fashion. I have gone through the famous cuda-convnet and also the recent OverFeat. While convnet is supervised, Overfeat has predefined weights learnt on Imagenet dataset.

Are there any implementations which will allow me to learn features in a totally unsupervised way from images? [Am I correct regarding convnet and overfeat? Or can they be used to learn hierarchies of features from images?]

Thanks :)

asked Jan 29 '14 at 03:39

Sharath Chandra
311●13●16●21

One Answer:

oldest newest most voted

The two main ones I know of are RBM and autoencoder (e.g. denoising, contractive).

answered Feb 19 '14 at 23:17

Ng0323
156●7●9●15

deepnet module is also another implementation of many deep architectures on cuda: https://github.com/nitishsrivastava/deepnet

(Feb 23 '14 at 23:03) Sharath Chandra

Subscription:

Once you sign in you will be able to subscribe for any updates here

Tags:

rbm ×121
unsupervised-learning ×40
image ×15
convolution-neural-networks ×10
gpu ×7
hierarchical ×6
cuda ×4

Asked: Jan 29 '14 at 03:39

Seen: 1,971 times

Last updated: Feb 23 '14 at 23:03

Unsupervised Feature Learning from Images on GPU

Subscription:

Related questions