Sparse coding in Python [closed]

Closed. This question is seeking recommendations for books, tools, software libraries, and more. It does not meet Stack Overflow guidelines. It is not currently accepting answers.

---

We don’t allow questions seeking recommendations for books, tools, software libraries, and more. You can edit the question so it can be answered with facts and citations.

Closed 6 years ago.

开发者_StackOverflow Improve this question

I'm looking for a library which implements the most common sparse coding and dictionary learning algorithms with a python interface, any suggestion?

---

Regarding this question; seems that a library which implements most of Sparse Coding algorithms is SPAMS http://spams-devel.gforge.inria.fr/ which now offers Python support

---

That's a pretty broad question. Among sparse coding, dictionary learning, and nonnegative factorization, there are plenty of methods, many of which are still so new that there does not yet exist a dominant piece of Python software that is universally accepted.

These research areas have been my recent focus, and I primarily use Python. I just decided to write my own simple versions of matching pursuit, NMF (and nonnegative LS), KSVD, and more. I might recommend doing the same for your specific purposes; they are not difficult to implement and modify. Just get used to Numpy, Scipy, and numpy.linalg.

That said, there are publicly available Python tools, but you will need to search more specifically. For instance, "matching pursuit python" yields a few results.

For those with convex optimization formulations, you could use CVX which is really easy to use, but you may not find it fast enough.

---

I guess these are answers if (a) you didn't find them and (b) there's nothing wrong with them.

http://www.eecs.berkeley.edu/~jiayq/sparsecoding.html

http://www.google-melange.com/gsoc/project/google/gsoc2011/vladn/11001

---

I use MPTK ("the Matching Pursuit Toolkit") - http://gforge.inria.fr/projects/mptk/ - there are some python wrappers in the code, but I don't know how stable they are. Personally I just use Python wrappers around the commandline utilities, because in my case the data is usually on disk rather than in memory so there's no big drawback to the loose coupling.