search_monte_carlo.hpp

Go to the documentation of this file.

#ifndef FSTSEARCHMONTECARLO_H
#define FSTSEARCHMONTECARLO_H
 /* 
=========================================================================
Copyright and license:
  * FST4 software (with the exception of any bundled or externally linked
    third-party components) is Copyright (C) 2000-present Institute of
    Information Theory and Automation of the Czech Academy of Sciences (UTIA).
  * FST4 is free software: you can redistribute it and/or modify it under
    the terms of the GNU Affero General Public License as published by the
    Free Software Foundation, version 3. The full text of the license is in
    the file LICENSE in the distribution root and at
    https://www.gnu.org/licenses/agpl-3.0.html.
  * FST4 is distributed in the hope that it will be useful, but WITHOUT ANY
    WARRANTY; without even the implied warranty of MERCHANTABILITY or
    FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public
    License for more details.
  * Closed-source use without the AGPL's copyleft obligations requires a
    commercial license from UTIA; in case of interest write to
    utia@utia.cas.cz. LICENSING.md in the distribution root explains both
    options, the contributor license agreement that every contribution
    requires, and the contacts (technical questions and pull requests:
    somol@utia.cas.cz, with "FST" in the subject line).
========================================================================== */
 
#include <memory>
#include <cassert>
#include <iostream>
#include <sstream>
#include <ctime>
#include "error.hpp"
#include "global.hpp"
#include "stopwatch.hpp"
#include "search.hpp"
#include "result_tracker.hpp"
#include "candidate_evaluator.hpp"
#include "candidate_evaluator_sequential.hpp"
 
/*============== Template parameter type naming conventions ==============
--------- Numeric types: -------------------------------------------------
DATATYPE - data sample values - usually real numbers (but may be integers
          in text processing etc.)
REALTYPE - must be real numbers - for representing intermediate results of 
          calculations like mean, covariance etc.
IDXTYPE - index values for enumeration of data samples - (nonnegative) integers, 
          extent depends on numbers of samples in data
DIMTYPE - index values for enumeration of features (dimensions), or classes (not 
          class sizes) - (nonnegative) integers, usually lower extent than IDXTYPE, 
          but be aware of expressions like _classes*_features*_features ! 
          in linearized representations of feature matrices for all classes
BINTYPE - feature selection marker type - represents ca. <10 different feature 
          states (selected, deselected, sel./desel. temporarily 1st nested loop, 2nd...)
RETURNTYPE - criterion value: real value, but may be extended in future to support 
          multiple values 
--------- Class types: ---------------------------------------------------
SUBSET       - class of class type Subset 
CLASSIFIER   - class implementing interface defined in abstract class Classifier 
EVALUATOR    - class implementing interface defined in abstract class Sequential_Step 
DISTANCE     - class implementing interface defined in abstract class Distance 
DATAACCESSOR - class implementing interface defined in abstract class Data_Accessor 
INTERVALCONTAINER - class of class type Data_Intervaller 
CONTAINER    - STL container of class type Data_Interval  
========================================================================== */
 
namespace FST {

 
template<class RETURNTYPE, typename DIMTYPE, class SUBSET, class CRITERION>
class Search_Monte_Carlo : public Search<RETURNTYPE,DIMTYPE,SUBSET,CRITERION>{
public:
        typedef Search<RETURNTYPE,DIMTYPE,SUBSET,CRITERION> parent; 
        typedef std::shared_ptr<CRITERION> PCriterion; 
        typedef std::shared_ptr<SUBSET> PSubset; 
        typedef Candidate_Evaluator<RETURNTYPE,DIMTYPE,SUBSET,CRITERION> EVALUATORENGINE; 
        typedef std::shared_ptr<EVALUATORENGINE> PEvaluatorEngine; 
        Search_Monte_Carlo(const unsigned long long trials_limit=1000, const unsigned long time_limit=0, const unsigned int time_limit_check_freq=1, PEvaluatorEngine engine=PEvaluatorEngine()):Search<RETURNTYPE,DIMTYPE,SUBSET,CRITERION>(),_engine(engine) {
                if(!_engine) _engine.reset(new Candidate_Evaluator_Sequential<RETURNTYPE,DIMTYPE,SUBSET,CRITERION>());
                notify("Search_Monte_Carlo constructor."); set_stopping_condition(trials_limit,time_limit,time_limit_check_freq); reset_cardinality_randomization();
        }
        virtual ~Search_Monte_Carlo() {notify("Search_Monte_Carlo destructor.");}
 
        // NOTE: crit is expected to be already initialize()d before use here
        virtual bool search(const DIMTYPE target_d, RETURNTYPE &result, const PSubset sub, const PCriterion crit, std::ostream& os=std::cout) override;

        void set_stopping_condition(const unsigned long long trials_limit=1000, const unsigned long time_limit=0, const unsigned int time_limit_check_freq=100) {if(trials_limit==0 && time_limit==0) {_trials_limit=1000; _time_limit=0;} else {_trials_limit=trials_limit; _time_limit=time_limit;} if(time_limit_check_freq==0) _time_limit_check_freq=100; else _time_limit_check_freq=time_limit_check_freq; }
 
        // set the way of random subset (size) generation
        void set_cardinality_randomization(const DIMTYPE d_from, const DIMTYPE d_to) {if(d_from>0 && d_from<=d_to) {_d_from=d_from; _d_to=d_to;} else {_d_from=_d_to=0;}} 
        void set_cardinality_randomization(const float d_prob) {if(d_prob>0.0 && d_prob<1.0) _d_prob=d_prob; else _d_prob=0.0;} 
        void reset_cardinality_randomization() {_d_from=_d_to=0; _d_prob=0.0;} 
 
        virtual std::ostream& print(std::ostream& os) const override;
 
protected:
        PEvaluatorEngine _engine; 
        unsigned long _time_limit; 
        unsigned long long _trials_limit; 
        unsigned int _time_limit_check_freq; 
        DIMTYPE _d_from; 
        DIMTYPE _d_to; 
        float _d_prob; 
};
 
template<class RETURNTYPE, typename DIMTYPE, class SUBSET, class CRITERION>
std::ostream& Search_Monte_Carlo<RETURNTYPE,DIMTYPE,SUBSET,CRITERION>::print(std::ostream& os) const
{
        if(_engine->is_threaded()) {
                os << "Random search  [Search_Monte_Carlo(max_threads="<<_engine->get_max_threads();
                if(_trials_limit>0) os << ", maxtrials="<<_trials_limit;
                if(_time_limit>0) os<<", seconds="<<_time_limit;
                if(_d_prob>0.0) os<<", d_prob="<<_d_prob;
                else os<<", d_from="<<_d_from<<", d_to="<<_d_to;
                os<<")";
        } else {
                os << "Random search  [Search_Monte_Carlo(";
                if(_trials_limit>0) os << "maxtrials="<<_trials_limit<<", ";
                if(_time_limit>0) os<<"seconds="<<_time_limit<<", ";
                if(_d_prob>0.0) os<<"d_prob="<<_d_prob;
                else os<<"d_from="<<_d_from<<", d_to="<<_d_to;
                os<<")";
        }
        if(parent::result_tracker_active()) os << " with " << *parent::_tracker;
        os<<"]";
        return os;
}
 
template<class RETURNTYPE, typename DIMTYPE, class SUBSET, class CRITERION>
bool Search_Monte_Carlo<RETURNTYPE,DIMTYPE,SUBSET,CRITERION>::search(const DIMTYPE target_d, RETURNTYPE &result, const PSubset sub, const PCriterion crit, std::ostream& os) // returns found subset + criterion value
{
        StopWatch swatch;
        assert(sub);
        assert(crit);
        assert(_engine);
        const DIMTYPE n=sub->get_n();
        assert(target_d<n);
 
        bool track=parent::result_tracker_active(); if(track) parent::_tracker->set_output_detail(parent::get_output_detail());
 
        if(parent::output_normal()) {
                msgbuf sos;
                sos << "---------------------------------------" << std::endl;
                sos << "Starting " << *this << std::endl;
                sos << "with Criterion: " << *crit << std::endl;
                if(target_d==0) sos << "Subset size to be optimized." << std::endl; else sos << "Target subset size set to: " << target_d << std::endl;
                sos << std::endl << std::flush;
                this->report(os,sos);
        }
        assert(_trials_limit!=0 || _time_limit!=0);
        assert(_time_limit_check_freq>0);
        unsigned long long _trials=0; // the printed twin of the engine's generation ordinal - same 64-bit domain
        unsigned long _time=0;
        // the random subset size range used when target_d==0 and no inclusion probability is set:
        // the configured [_d_from,_d_to] clamped into [1,n]; with no range configured it is the full [1,n]
        DIMTYPE _d_from_tmp=_d_from, _d_to_tmp=_d_to;
        if(_d_to_tmp<1 || _d_to_tmp>n) _d_to_tmp=n;
        if(_d_from_tmp<1) _d_from_tmp=1; else if(_d_from_tmp>_d_to_tmp) _d_from_tmp=_d_to_tmp;
        assert(_d_from_tmp>=1 && _d_from_tmp<=_d_to_tmp && _d_to_tmp<=n);
 
        // random candidates are generated until the stopping condition (trials and/or
        // time limit) is met; both engines call the generator on the MASTER thread only,
        // so the random draws all come from one fst_rand stream (same seed = same
        // candidate sequence, threaded or not) and the trial counter needs no synchronization
        StopWatch *psw=&swatch;
        typename EVALUATORENGINE::Generator generator=[this,psw,&_trials,&_time,_d_from_tmp,_d_to_tmp,target_d](SUBSET &cand)->bool {
                if( (_trials_limit!=0 && _trials>=_trials_limit) || (_time_limit!=0 && _time>=_time_limit) ) return false;
                if(target_d==0) {
                        if(_d_prob>0.0) do {cand.make_random_subset(_d_prob);} while(cand.get_d()==0);
                        else cand.make_random_subset(_d_from_tmp,_d_to_tmp);
                } else cand.make_random_subset(target_d);
                ++_trials;
                if(_trials%_time_limit_check_freq==0) _time=(unsigned long)psw->time_elapsed();
                return true;
        };
 
        typename EVALUATORENGINE::BetterPredicate better=[](const RETURNTYPE val, const PSubset &cand, const RETURNTYPE bestval, const PSubset &bestsub)->bool {
                return val>bestval || (val==bestval && cand->get_d()<bestsub->get_d());
        };
 
        typename EVALUATORENGINE::BestCallback on_best;
        if(parent::output_normal()) {
                std::ostream *pos=&os;
                // both engines fire this on the master in generation order (tidx==-1) with the
                // candidate's generation ordinal == the trial index, threaded or not
                on_best=[this,psw,pos](const int /*tidx*/, const unsigned long long ordinal, const RETURNTYPE val, const PSubset &bsub) {
                        msgbuf sos;
                        sos << "new MAXCRIT="<<val<< ", trial=" << ordinal << ", " << *bsub << std::endl << *psw << std::endl << std::endl << std::flush;
                        this->report_solution(*pos,val,*bsub,sos);
                };
        }
 
        if(!_engine->evaluate_stream(generator, crit, track? parent::_tracker : typename parent::PResultTracker(), better, on_best, result, sub)) return false;
 
        if(parent::output_normal()) {
                msgbuf sos;
                sos << "......................................." << std::endl;
                if(_engine->is_threaded()) sos << "Search_Monte_Carlo_Threaded() search finished after "<<_trials<<" trials. " << swatch << std::endl;
                else sos << "Search_Monte_Carlo() search finished after "<<_trials<<" trials. " << swatch << std::endl;
                sos << "Search result: "<< std::endl << *sub << std::endl << "Criterion value: " << result << std::endl << std::endl << std::flush;
                this->report(os,sos);
        }
        return true;
}
 
} // namespace
#endif // FSTSEARCHMONTECARLO_H ///:~