multilabel_oaa.cc

Go to the documentation of this file.

/*
Copyright (c) by respective owners including Yahoo!, Microsoft, and
individual contributors. All rights reserved.  Released under a BSD (revised)
license as described in the file LICENSE.
 */
#include <sstream>
#include <float.h>
#include "reductions.h"
#include "vw.h"

using namespace VW::config;

struct multi_oaa
{
  size_t k;
};

template <bool is_learn>
void predict_or_learn(multi_oaa& o, LEARNER::single_learner& base, example& ec)
{
  MULTILABEL::labels multilabels = ec.l.multilabels;
  MULTILABEL::labels preds = ec.pred.multilabels;
  preds.label_v.clear();

  ec.l.simple = {FLT_MAX, 1.f, 0.f};
  uint32_t multilabel_index = 0;
  for (uint32_t i = 0; i < o.k; i++)
  {
    if (is_learn)
    {
      ec.l.simple.label = -1.f;
      if (multilabels.label_v.size() > multilabel_index && multilabels.label_v[multilabel_index] == i)
      {
        ec.l.simple.label = 1.f;
        multilabel_index++;
      }
      base.learn(ec, i);
    }
    else
      base.predict(ec, i);
    if (ec.pred.scalar > 0.)
      preds.label_v.push_back(i);
  }
  if (is_learn && multilabel_index < multilabels.label_v.size())
    std::cout << "label " << multilabels.label_v[multilabel_index] << " is not in {0," << o.k - 1
              << "} This won't work right." << std::endl;

  ec.pred.multilabels = preds;
  ec.l.multilabels = multilabels;
}

void finish_example(vw& all, multi_oaa&, example& ec)
{
  MULTILABEL::output_example(all, ec);
  VW::finish_example(all, ec);
}

LEARNER::base_learner* multilabel_oaa_setup(options_i& options, vw& all)
{
  auto data = scoped_calloc_or_throw<multi_oaa>();
  option_group_definition new_options("Multilabel One Against All");
  new_options.add(make_option("multilabel_oaa", data->k).keep().help("One-against-all multilabel with <k> labels"));
  options.add_and_parse(new_options);

  if (!options.was_supplied("multilabel_oaa"))
    return nullptr;

  LEARNER::learner<multi_oaa, example>& l = LEARNER::init_learner(data, as_singleline(setup_base(options, all)),
      predict_or_learn<true>, predict_or_learn<false>, data->k, prediction_type::multilabels);
  l.set_finish_example(finish_example);
  all.p->lp = MULTILABEL::multilabel;
  all.label_type = label_type::multi;
  all.delete_prediction = MULTILABEL::multilabel.delete_label;

  return make_base(l);
}