Revision history[back]
click to hide/show revision 1
Revision n. 1

Jun 15 '10 at 11:24

Joseph%20Turian's gravatar image

Joseph Turian
579051125146

How do I sparsify a dense feature representation?

Let's say I have a low-dimensional dense feature representation. I would like to transform this dense, low-dimensional representation into a sparse, high-dimensional representation, which can be better input for a linear classifier.

How do I sparsify a dense feature representation?

click to hide/show revision 2
Revision n. 2

Jun 15 '10 at 16:00

Joseph%20Turian's gravatar image

Joseph Turian
579051125146

How do I sparsify a dense feature representation?

Let's say I have a low-dimensional dense feature representation. I would like to transform this dense, low-dimensional representation into a sparse, high-dimensional representation, which can be better input for a linear classifier.

How do I sparsify a dense feature representation?

Here's a concrete example of how this could me used: Let's say I have 1M images, each of them 32x32 pixels. Instead of representing each image as a dense vector with 1024 entries, I would maybe want a 10K vector with only 10 non-zeroes.

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.