SharkBSGDTrain: Budgeted SGD Train.

In RcppShark: R Interface to the Shark Machine Learning Library

Description

This will train a support vector machine by an SGD (Pegasos-like). It will limit the number of support vector machines by applying the given budget maintenance strategy. See more on http://image.diku.dk/shark/sphinx_pages/build/html/rest_sources/tutorials/algorithms/kernelBudgetedSGD.html

Usage

SharkBSGDTrain(x, y = NULL, verbose = FALSE, budget = 500,
  strategy = "Merge", C = 1, gamma = 1, epochs = 1)

Arguments

x	matrix with input data
y	vector with labels
verbose	verbose output?
budget	size of budget
strategy	strategy to use to maintain the budget size. choises are 'Merge', 'RemoveSmallest', 'RemoveRandom, 'Project'
C	regularization constant
gamma	kernel bandwidth for RBF kernel
epochs	number of iterations through data set

Note

Currently works only for binary classification. Uses only RBF kernel.

RcppShark documentation built on May 30, 2017, 6:20 a.m.