Given a list of feature dictionaries (sparse feature vectors)
representing different objects and a matching list of target values to
predict, assign values to each feature encountered to best predict the
target values.
Ultimately, this essentially gets laid out as a system of linear
equations where the value of the feature vectors represents the
coefficients of the equations and the target values are the equation
solutions. The problem then is to just find the set of unknown variable
values which solves the system with the minimal error.
|
|
__init__(self)
Default constructor. |
|
|
|
|
main(self,
argv)
For use as command-line executable method |
|
|
|
|
regressionByFilenames(self,
featureDictFilename,
featureValueFilename,
targetValueFilename)
Simple wrapper around regression method, just to open up files |
|
|
|
|
writeFeatureValueDict(self,
featureValueDict,
featureValueFile)
Write out the contents of the featureValueDict to the
featureValueFile. |
|
|
|
|
predictionByFilenames(self,
featureDictFilename,
featureValueFilename,
targetValueFilename)
Simple wrapper around prediction method, just to open up files |
|
|
|
|
readFeatureValueDict(self,
featureValueFile)
Read the contents of the feature value file, probably generated by
this classes regression method, back into a dictionary. |
|
|
|
|
regression(self,
featureDictList,
targetValueList)
Primary regression method: Given a list over all of the input
object feature dictionaries (or just using a
FeatureDictReaderFactory) and a list of the targetValues, find the
optimal solution for the "value" of each feature
contributing to the total target values. |
|
|
|
|
prediction(self,
featureDictList,
featureValueDict)
Primary prediction method (reciprocal function of regression):
Given a list over all of the input object feature dictionaries and a
dictionary mapping features to their "value," generate
predicted values for the "total" targetValue of each input
feature dictionary. |
|
|
|
|
predictionSingle(self,
featureDict,
featureValueDict)
Generate a single prediction by going over all of the features in
the feature dictionary and looking up values to sum over in the
featureValueDict. |
|
|