3 feb 2008 kl. 06.49 skrev Ben Kucinich:
I am interested in this project as I have done a lot of AI
programming in
Java. I have used Hadoop but I have never programmed in it.
Therefore, my
questions are:-
1. Is the knowledge of Hadoop necessary to contribute to this project?
Absolutely not.
Yonik's Law of Patches says: A half-baked patch in Jira, with no
documentation, no tests and no backwards compatibility is better than
no patch at all.
2. Is there any scope for including graph search and path finding
algorithms
related to graph. If you need contributors for graph theory, I would
like to
contribute by writing utils classes for Graph search and finding
problems.
This is the first time someone mentions it on this forum. The semi-
official short term goals is to implement a few of the classifiers
(NB, NN, SVM) and clusterers (k-means, EM) algorithms in the paper <http://www.cs.stanford.edu/people/ang//papers/nips06-mapreducemulticore.pdf
>, but the project is in no way limited to this.
3. Is there a list anywhere in the wiki where I can see what
algorithms and
classes have to be written? Can I add the graph classes I have
thought of in
this.
The Stanford paper is our current guide, but it does not contain all
the things we have been speaking of on this forum. Feature selection,
HMM, Dijkstra and what not. I hope the Mahout Jira will be the Java
community de facto place to post ML-related things, Hadooped or not.
karl
