[mlpack-svn] [MLPACK] #337: Normalization for Collaborative Filtering
MLPACK Trac
trac at coffeetalk-1.cc.gatech.edu
Wed Feb 26 20:30:05 EST 2014
#337: Normalization for Collaborative Filtering
---------------------------+------------------------------------------------
Reporter: sumedhghaisas | Owner:
Type: defect | Status: new
Priority: major | Milestone: mlpack 1.0.9
Component: mlpack | Keywords: Normalization, Collaborative Filtering
Blocking: | Blocked By:
---------------------------+------------------------------------------------
Option added for collaborative filtering model to perform normalization
before applying ALS. The normalizations replaces all the missing entries
with column averages and then subtracts row averages from all the entries.
Laplace's Smoothing is added for smoothening the sparse rating matrix.
Files Included:
1) cf.cpp : Added default constructor parameter normalize. CleanData is
modified to
apply laplace's smoothing and compute comulmnAverages and
rowAverages.
Sparse matrix is not modified at this stage. The normalization
will be
while actually performing ALS (through NMF).
2) als_normalized_update_rule : WNormalizedAlternatingLeastSqureRule and
HNormalizedAlternatingLeastSquareRule
2) nmf.hpp : Added template specialized class to handle the added update
rules
3) nmf_impl.hpp : Implementation of new Apply method which takes column
average and row average matrices
--
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/337>
MLPACK <www.fast-lab.org>
MLPACK is an intuitive, fast, and scalable C++ machine learning library developed at Georgia Tech.
More information about the mlpack-svn
mailing list