classifier_extern.hpp

Go to the documentation of this file.

#ifndef FSTCLASSIFIEREXTERN_H
#define FSTCLASSIFIEREXTERN_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 <vector>
#include <string>
#include <cstdio>
#include <cstdlib>
#include <cassert>
#include <iostream>
#include "error.hpp"
#include "global.hpp"
#include "clonable.hpp"
#include "classifier.hpp"
#include "extern_bridge.hpp" // the platform-sensitive process plumbing, shared with Criterion_Extern_Value
 
/*============== 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 DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
class Classifier_Extern : public Clonable_Impl<Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>,Classifier<RETURNTYPE,DIMTYPE,SUBSET,DATAACCESSOR> > {
public:
        typedef Classifier<RETURNTYPE,DIMTYPE,SUBSET,DATAACCESSOR> parent; 
        typedef Clonable_Impl<Classifier_Extern,parent> parent_impl; 
        typedef std::shared_ptr<DATAACCESSOR> PDataAccessor;      
        typedef std::shared_ptr<SUBSET> PSubset;                  
        typedef typename DATAACCESSOR::PPattern PPattern;         
        Classifier_Extern() : _bridge("Classifier_Extern"), _classes(0), _features(0), _subfeatures(0) {notify("Classifier_Extern constructor.");}
        virtual ~Classifier_Extern() {notify("Classifier_Extern destructor.");} // the bridge destructor shuts the child down
 
        void set_command(const std::string &cmd) {_bridge.set_command(cmd);}       
        const std::string& get_command() const {return _bridge.get_command();}    

        void initialize(const PDataAccessor da);

        virtual bool classify(DIMTYPE &cls, const PPattern &pattern) override;
        virtual bool train(const PDataAccessor da, const PSubset sub) override;
        virtual bool test(RETURNTYPE &result, const PDataAccessor da) override;
        virtual bool test_confusion(std::vector<RETURNTYPE> &confusion, DIMTYPE &classes, const PDataAccessor da) override;
 
        virtual std::ostream& print(std::ostream& os) const override {os << "Classifier_Extern(cmd=\"" << _bridge.get_command() << "\")"; return os;}
private:
        friend parent_impl; 
        Classifier_Extern(const Classifier_Extern& ce) :
                _bridge("Classifier_Extern"), // clones NEVER share pipes — each fresh bridge spawns its own child lazily
                _classes(ce._classes),
                _features(ce._features),
                _subfeatures(ce._subfeatures),
                _index(ce._index)
        {
                _bridge.set_command(ce._bridge.get_command());
                notify("Classifier_Extern copy-constructor.");
        }
protected:
        Extern_Bridge _bridge;         
        DIMTYPE _classes;              
        DIMTYPE _features;             
        DIMTYPE _subfeatures;          
        std::vector<DIMTYPE> _index;   

        void send_row(const DIMTYPE label_or_none, const bool with_label, const PPattern &pattern);
        long parse_prediction(const std::string &line);
        bool run_test_block(std::vector<DIMTYPE> &predictions, const PDataAccessor da);
};
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
void Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::initialize(const PDataAccessor da)
{
        notify("Classifier_Extern::initialize().");
        assert(da);
        assert(da->getNoOfFeatures()>0);
        if(da->getNoOfClasses()<2) throw fst_error("Classifier_Extern requires at least 2 classes");
        if(_bridge.get_command().empty()) throw fst_error("Classifier_Extern: no command set (see --extern-cmd)");
        _classes=da->getNoOfClasses();
        _features=da->getNoOfFeatures();
        _index.resize(_features);
        _subfeatures=0; // no model yet
}
 
// (the process spawn/shutdown/line-I/O machinery lives in extern_bridge.hpp, shared with
//  Criterion_Extern_Value; its error messages are prefixed with the component name)
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
void Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::send_row(const DIMTYPE label_or_none, const bool with_label, const PPattern &pattern)
{
        std::FILE *out=_bridge.to_child();
        char buf[64];
        if(with_label) {std::snprintf(buf,sizeof(buf),"%ld",(long)label_or_none); std::fputs(buf,out);}
        for(DIMTYPE f=0;f<_subfeatures;f++)
        {
                std::snprintf(buf,sizeof(buf),"%.17g",(double)pattern[_index[f]]); // bit round-trip for doubles
                if(with_label || f>0) std::fputc(' ',out);
                std::fputs(buf,out);
        }
        if(std::fputc('\n',out)==EOF)
                _bridge.die_on_write();
}
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
long Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::parse_prediction(const std::string &line)
{
        // strict full-line numeric parse: a bare strtol would accept any garbage line (a stray child
        // diagnostic, an "ERROR ..." message) as a VALID class 0 prediction and silently produce
        // wrong results; the TRAIN path fails loudly on non-OK, TEST must be equally loud
        const char *s=line.c_str();
        char *end=0;
        const long v=std::strtol(s,&end,10);
        if(end==s || *end!='\0')
                {_bridge.shutdown_process(); throw fst_error("Classifier_Extern: '"+_bridge.get_command()+"' returned a non-numeric prediction line: '"+line+"'");}
        return v;
}
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
bool Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::train(const PDataAccessor da, const PSubset sub)
{
        notify("Classifier_Extern::train().");
        assert(da);
        assert(sub);
        assert(_classes==da->getNoOfClasses());
        assert(_features==da->getNoOfFeatures());
        assert(sub->get_n_raw()==da->getNoOfFeatures());
 
        // buffer the feature subset index
        _subfeatures=0;
        for(DIMTYPE fsel : sub->selected_features()) {assert(_subfeatures<_features); _index[_subfeatures++]=fsel;}
        assert(_subfeatures==sub->get_d_raw());
        assert(_subfeatures>0);
 
        _bridge.ensure_process();
 
        // header first: count the training patterns (cheap block-size sweep)
        unsigned long long n=0;
        {typename DATAACCESSOR::PPattern p; IDXTYPE s,i; (void)p; (void)i;
         for(DIMTYPE c=0;c<_classes;c++) {da->setClass(c); for(const auto &blk : da->blocks(TRAIN)) {s=blk.size; n+=(unsigned long long)s;}}}
        if(n==0) return false;
        _bridge.send_line("TRAIN "+std::to_string(n)+" "+std::to_string((long)_subfeatures)+" "+std::to_string((long)_classes));
 
        typename DATAACCESSOR::PPattern p;
        IDXTYPE s,i;
        const DIMTYPE stride=da->getNoOfFeatures();
        for(DIMTYPE c=0;c<_classes;c++)
        {
                da->setClass(c);
                for(const auto &blk : da->blocks(TRAIN)) for(i=0,p=blk.data,s=blk.size;i<s;i++)
                        send_row(c,true,&p[(std::size_t)i*stride]);
        }
        _bridge.flush_to_child();
        const std::string resp=_bridge.read_line();
        if(resp!="OK") throw fst_error("Classifier_Extern: '"+_bridge.get_command()+"' rejected TRAIN: "+resp);
        return true;
}
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
bool Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::classify(DIMTYPE &cls, const PPattern &pattern)
{
        assert(_subfeatures>0); // train() must have succeeded
        _bridge.ensure_process();
        _bridge.send_line("TEST 1 "+std::to_string((long)_subfeatures));
        send_row(0,false,pattern);
        _bridge.flush_to_child();
        const long v=parse_prediction(_bridge.read_line());
        if(v<0 || v>=(long)_classes) {_bridge.shutdown_process(); return false;} // desynced/misbehaving child: respawn on next use
        cls=(DIMTYPE)v;
        return true;
}
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
bool Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::run_test_block(std::vector<DIMTYPE> &predictions, const PDataAccessor da)
{
        // one batched TEST block over the designated test part; prediction order =
        // the canonical enumeration (class-major, block order) shared by all classifiers
        assert(da);
        assert(_subfeatures>0);
        _bridge.ensure_process();
        unsigned long long m=0;
        {IDXTYPE s; for(DIMTYPE c=0;c<_classes;c++) {da->setClass(c); for(const auto &blk : da->blocks(TEST)) {s=blk.size; m+=(unsigned long long)s;}}}
        if(m==0) return false;
        _bridge.send_line("TEST "+std::to_string(m)+" "+std::to_string((long)_subfeatures));
        typename DATAACCESSOR::PPattern p;
        IDXTYPE s,i;
        const DIMTYPE stride=da->getNoOfFeatures();
        for(DIMTYPE c=0;c<_classes;c++)
        {
                da->setClass(c);
                for(const auto &blk : da->blocks(TEST)) for(i=0,p=blk.data,s=blk.size;i<s;i++)
                        send_row(0,false,&p[(std::size_t)i*stride]);
        }
        _bridge.flush_to_child();
        predictions.clear();
        predictions.reserve((std::size_t)m);
        for(unsigned long long k=0;k<m;k++)
        {
                const long v=parse_prediction(_bridge.read_line());
                if(v<0 || v>=(long)_classes) {
                        // the remaining m-k-1 response lines stay unread - a desynced child would make
                        // the NEXT round-trip fail with a misleading "rejected TRAIN: <number>" error;
                        // drop the child instead, the next evaluation respawns it cleanly
                        _bridge.shutdown_process();
                        return false;
                }
                predictions.push_back((DIMTYPE)v);
        }
        return true;
}
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
bool Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::test(RETURNTYPE &result, const PDataAccessor da)
{
        notify("Classifier_Extern::test().");
        std::vector<DIMTYPE> pred;
        if(!run_test_block(pred,da)) return false;
        // tally against true labels in the same canonical enumeration order
        unsigned long long count=0, correct=0;
        IDXTYPE s,i;
        for(DIMTYPE c_test=0;c_test<_classes;c_test++)
        {
                da->setClass(c_test);
                for(const auto &blk : da->blocks(TEST)) for(i=0,s=blk.size;i<s;i++)
                {
                        if(pred[count]==c_test) correct++;
                        count++;
                }
        }
        assert(count==pred.size());
        if(count==0) return false;
        result=(RETURNTYPE)correct/(RETURNTYPE)count;
        return true;
}
 
template<class RETURNTYPE, typename DATATYPE, typename REALTYPE, typename IDXTYPE, typename DIMTYPE, class SUBSET, class DATAACCESSOR>
bool Classifier_Extern<RETURNTYPE,DATATYPE,REALTYPE,IDXTYPE,DIMTYPE,SUBSET,DATAACCESSOR>::test_confusion(std::vector<RETURNTYPE> &confusion, DIMTYPE &classes, const PDataAccessor da)
{
        // batched override: ONE TEST round-trip instead of the base default's per-pattern classify()
        notify("Classifier_Extern::test_confusion().");
        std::vector<DIMTYPE> pred;
        if(!run_test_block(pred,da)) return false;
        classes=_classes;
        confusion.assign((std::size_t)_classes*(std::size_t)_classes,(RETURNTYPE)0.0);
        unsigned long long count=0;
        IDXTYPE s,i;
        for(DIMTYPE c_test=0;c_test<_classes;c_test++)
        {
                da->setClass(c_test);
                for(const auto &blk : da->blocks(TEST)) for(i=0,s=blk.size;i<s;i++)
                {
                        confusion[(std::size_t)c_test*(std::size_t)_classes+(std::size_t)pred[count]]+=1.0;
                        count++;
                }
        }
        assert(count==pred.size());
        return count>0;
}
 
} // namespace
#endif // FSTCLASSIFIEREXTERN_H ///:~