All the information is given about the question. I am learning IRIS recognition and during this i got Sparse filtering.
Asked
Active
Viewed 1,137 times
1 Answers
2
l2 ball -
suppose a vector x = 2x + 2y + 2z then l2 norm of this vector is square root of (sqr(2x) + sqr(2y) + sqr(2z)). Now when we find f(x) / (l2-norm of f(x)), this is equivalent to unit vector. Now, here f(x) is feature of one example, like-wise all features by this way lie on a sphere(circle if 2-D) of radius one. So, this is l2- ball.
Now Question is What is l1-penalty ? I hope, we will get it soon.

Abhishek Raj
- 478
- 7
- 17