|
This provides code for working with support vector machines.
Support vector machine (SVM) is a supervised machine learning
algorithm that has been shown to work well on a variety of
classification problems.
XXX describe kernel_fn, C, epsilon
For more information, see:
http://svm.first.gmd.de/
http://svm.research.bell-labs.com/
Classes:
SVM Holds data for a Support Vector Machine.
SMOTrainer Trains an SVM using Sequential Minimal Optimization (SMO).
KeerthiTrainer Trains an SVM using Keerthi's extensions to SMO.
TransductiveTrainer Trains a transductive SVM.
LinearKernel
PolynomialKernel
RadialBasisFunctionKernel
HyperbolicTangentKernel
Functions:
train Train a Support Vector Machine on some training data.
trans_transductive
classify Use a Support Vector Machine to classify some data.
Usage:
The train function is provided as a user-friendly interface module.
Use this function if all you want is a plain-vanilla SVM classifier.
However, if you're concerned about details such as the training
algorithm, you will need to use some of the classes provided.
|
Functions
|
|
_dot
_sign
_sparse_dot
_sparse_subtract
_subtract
classify
train
train_transductive
|
|
|
_dot
|
_dot ( x, y )
_dot(x, y) -> x*y
Return the dot product of x and y, where x and y are lists of numbers.
|
Exceptions
|
|
ValueError, "vectors must be same length"
|
|
|
|
_sign
|
_sign ( x )
_sign(x) -> 1 or -1
Return 1/-1 depending on the sign of x.
|
|
|
_sparse_dot
|
_sparse_dot ( x, y )
|
|
|
_sparse_subtract
|
_sparse_subtract ( x, y )
|
|
|
_subtract
|
_subtract ( x, y )
_subtract(x, y) -> x-y
Return x subtract y, where x and y are lists of numbers.
|
Exceptions
|
|
ValueError, "vectors must be same length"
|
|
|
|
classify
|
classify ( svm, x )
classify(svm, x) -> score
Classify x based on an SVM object.
|
|
|
train
|
train (
training_set,
results,
kernel_fn=LinearKernel(),
C=1.0,
epsilon=1E-3,
update_fn=None,
)
train(training_set, results,
kernel_fn=LinearKernel(), C=1.0, epsilon=1E-3, update_fn=None) -> SVM
Train a new support vector machine. training_set is a list
containing a list of numbers, each which is a vector in the
training set. results is a parallel list that contains either
1 or -1, which describes the class that the training vector belongs
to.
kernel_fn, C, and epsilon are optional parameters used to tune the
support vector machine.
update_fn is an optional callback function that called at the
end of every round of training. It should take 2 parameters:
update_fn(num_changed, svm)
|
|
|
train_transductive
|
train_transductive (
training_set,
results,
test_set,
num_pos,
kernel_fn=LinearKernel(),
C=1.0,
C_test=0.5,
epsilon=1E-3,
update_fn=None,
)
train_transductive(training_set, results, test_set, num_pos,
kernel_fn=LinearKernel(), C=1.0, C_test=0.5,
epsilon=1E-3, update_fn=None) -> SVM
XXX document this
|
|
Classes
|
|
|
|
|