it.uniroma2.sag.kelp.learningalgorithm.budgetedAlgorithm

Class RandomizedBudgetPerceptron

java.lang.Object

it.uniroma2.sag.kelp.learningalgorithm.budgetedAlgorithm.BudgetedLearningAlgorithm

it.uniroma2.sag.kelp.learningalgorithm.budgetedAlgorithm.RandomizedBudgetPerceptron

All Implemented Interfaces:

BinaryLearningAlgorithm, LearningAlgorithm, MetaLearningAlgorithm, OnlineLearningAlgorithm

public class RandomizedBudgetPerceptron
extends BudgetedLearningAlgorithm

It is a variation of the Randomized Budget Perceptron proposed in

[CavallantiCOLT2006] G. Cavallanti, N. Cesa-Bianchi, C. Gentile. Tracking the best hyperplane with a simple budget Perceptron. In proc. of the 19-th annual conference on Computational Learning Theory. (2006)

Until the budget is not reached the online learning updating policy is the one of the baseAlgorithm that this meta-algorithm is exploiting. When the budget is full, a random support vector is deleted and the perceptron updating policy is adopted

Author:

Simone Filice

Field Summary

Fields inherited from class it.uniroma2.sag.kelp.learningalgorithm.budgetedAlgorithm.BudgetedLearningAlgorithm

baseAlgorithm, budget, label

Constructor Summary

Constructors

Constructor and Description
RandomizedBudgetPerceptron()
RandomizedBudgetPerceptron(int budget, OnlineLearningAlgorithm baseAlgorithm, long seed, List<Label> labels)

Method Summary

Methods

Modifier and Type	Method and Description
RandomizedBudgetPerceptron	duplicate() Creates a new instance of the LearningAlgorithm initialized with the same parameters of the learningAlgorithm to be duplicated.
Prediction	predictAndLearnWithFullBudget(Example example) Learns from a single example applying a specific policy that must be adopted when the budget is reached
void	reset() Resets all the learning process, returning to the default state.
void	setSeed(long seed) Sets the seed for the random generator adopted to select the support vector to delete

Methods inherited from class it.uniroma2.sag.kelp.learningalgorithm.budgetedAlgorithm.BudgetedLearningAlgorithm

getBaseAlgorithm, getBudget, getLabel, getLabels, getPredictionFunction, learn, learn, setBaseAlgorithm, setBudget, setLabel, setLabels

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

RandomizedBudgetPerceptron

public RandomizedBudgetPerceptron()

RandomizedBudgetPerceptron

public RandomizedBudgetPerceptron(int budget,
                          OnlineLearningAlgorithm baseAlgorithm,
                          long seed,
                          List<Label> labels)

Method Detail

setSeed

public void setSeed(long seed)

Sets the seed for the random generator adopted to select the support vector to delete

Parameters:

seed - the seed of the randomGenerator

duplicate

public RandomizedBudgetPerceptron duplicate()

Description copied from interface: LearningAlgorithm

Creates a new instance of the LearningAlgorithm initialized with the same parameters of the learningAlgorithm to be duplicated. A shallow copy of the learning parameters is performed, while the state parameters are initialized to their default value. The labels are not initialized neither copied

reset

public void reset()

Description copied from interface: LearningAlgorithm

Resets all the learning process, returning to the default state. All the learned process is forgotten

predictAndLearnWithFullBudget

public Prediction predictAndLearnWithFullBudget(Example example)

Description copied from class: BudgetedLearningAlgorithm

Learns from a single example applying a specific policy that must be adopted when the budget is reached

Specified by:

predictAndLearnWithFullBudget in class BudgetedLearningAlgorithm

Parameters:

example - the example to be exploited in the learning process

Returns:

the prediction on the given example, using the old classification function (before any possible update)