provide hooks for alternative weight update procedures

[anoopsarkar]

for instance we should have hooks for the following improvements to the perceptron:

• [x] averaged perceptron
• [ ] MIRA (margin infused perceptron updates) or PA updates
• [ ] AdaGrad (Duchi et al)
• [ ] SAG (Schmidt et al)

[anoopsarkar]

For the pseudo-code of averaged perceptron, read:

• Michael Collins. Ranking algorithms for named entity extraction: Boosting and the voted perceptron. ACL 2002. http://aclweb.org/anthology/P/P02/P02-1062.pdf
• Chapter 2 of Dong's thesis: http://www.cs.sfu.ca/~anoop/students/dsong/dsong_msc_thesis.pdf

[yulanh]

Implemented average perceptron.

TODO: Speed up by hashing the features and using sparse matrix in scipy

[anoopsarkar]

Let us implement MIRA next. This update procedure is explained in Ryan McDonald's thesis.

[anoopsarkar]

closing this and adding a new issue for MIRA implementation.