SVRGMinimizer.h

Go to the documentation of this file.

/*
 * Copyright (c) The Shogun Machine Learning Toolbox
 * Written (w) 2015 Wu Lin
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * 1. Redistributions of source code must retain the above copyright notice, this
 *    list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright notice,
 *    this list of conditions and the following disclaimer in the documentation
 *    and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * The views and conclusions contained in the software and documentation are those
 * of the authors and should not be interpreted as representing official policies,
 * either expressed or implied, of the Shogun Development Team.
 *
 */

#ifndef SVRGMINIMIZER_H
#define SVRGMINIMIZER_H
#include <shogun/optimization/FirstOrderStochasticMinimizer.h>
#include <shogun/optimization/FirstOrderSAGCostFunction.h>
namespace shogun
{

class SVRGMinimizer: public FirstOrderStochasticMinimizer
{
public:
    SVRGMinimizer();

    SVRGMinimizer(FirstOrderSAGCostFunction *fun);

    virtual ~SVRGMinimizer();

    virtual float64_t minimize();

    virtual const char* get_name() const { return "SVRGMinimizer"; }


    virtual void set_sgd_number_passes(int32_t sgd_passes)
    {
        REQUIRE(m_num_passes>0, "Must set num_passes first\n");
        REQUIRE(sgd_passes>=0, "The number (%d) to go through data using SGD update must be positive\n",
            sgd_passes);
        REQUIRE(m_num_passes>sgd_passes,
            "SVRG update is not actived because the total number (%d) to go through data",
            " is less than the number (%d) to go through data using SGD\n",
            m_num_passes, sgd_passes);
        m_num_sgd_passes=sgd_passes;
    }

    virtual void set_average_update_interval(int32_t interval)
    {
        REQUIRE(m_num_passes>0, "Must set num passes first\n");
        REQUIRE(m_num_sgd_passes>=0, "Must set sgd update passes first\n");
        REQUIRE(interval>0, "Interval (%d) must be positive\n", interval);
        REQUIRE((m_num_passes-m_num_sgd_passes)%interval==0, "Interval is not valid\n");
        /* if (m_num_passes-m_num_sgd_passes)%interval!=0, will affect the finaly result if we do the following operations:
        * first do minimization, then save_to_file, and then load_from_file and finaly do minimization.
        * If we want to get the exact result when (m_num_passes-m_num_sgd_passes)%interval!=0,
        * we should store/restore m_average_gradient and m_previous_variable in save_to_file/load_from_file
        */
        m_svrg_interval=interval;
    }

protected:
    virtual void init_minimization();

    int32_t m_num_sgd_passes;

    int32_t m_svrg_interval;

    SGVector<float64_t> m_average_gradient;

    SGVector<float64_t> m_previous_variable;
private:
    void init();
};

}
#endif /* SVRGMINIMIZER_H */