scw 1.0

An implementation of Exact Soft Confidence-Weighted Learning

The explanation of the algorithm

This is an online supervised learning algorithm.

This learning method enjoys all the four salient properties:

• Large margin training

• Confidence weighting

• Capability to handle non-separable data

• Adaptive margin

The paper is here.

There are 2 kinds of implementations presented in the paper, which served as

scw.SCW1(C, ETA)
scw.SCW2(C, ETA)

in the code. C and ETA are hyperparameters.

Usage

from scw import SCW1, SCW2

scw = SCW1(C=1.0, ETA=1.0)
weights, covariance = scw.fit(training_data, teachers)
results = scw.perdict(test_data)

teachers is 1-dimensional and training_data and test_data are 2-dimensional array.