libopenms-dev 1.11.1-5 / usr / include / OpenMS / ANALYSIS / ID / HiddenMarkovModel.h

// --------------------------------------------------------------------------
//                   OpenMS -- Open-Source Mass Spectrometry
// --------------------------------------------------------------------------
// Copyright The OpenMS Team -- Eberhard Karls University Tuebingen,
// ETH Zurich, and Freie Universitaet Berlin 2002-2013.
//
// This software is released under a three-clause BSD license:
//  * Redistributions of source code must retain the above copyright
//    notice, this list of conditions and the following disclaimer.
//  * 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.
//  * Neither the name of any author or any participating institution
//    may be used to endorse or promote products derived from this software
//    without specific prior written permission.
// For a full list of authors, refer to the file AUTHORS.
// --------------------------------------------------------------------------
// 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 ANY OF THE AUTHORS OR THE CONTRIBUTING
// INSTITUTIONS 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.
//
// --------------------------------------------------------------------------
// $Maintainer: Andreas Bertsch $
// $Authors: Andreas Bertsch $
// --------------------------------------------------------------------------


#ifndef OPENMS_ANALYSIS_ID_HIDDENMARKOVMODEL_H
#define OPENMS_ANALYSIS_ID_HIDDENMARKOVMODEL_H

#include <vector>
#include <set>

#include <OpenMS/DATASTRUCTURES/Map.h>
#include <OpenMS/CONCEPT/Types.h>
#include <OpenMS/DATASTRUCTURES/StringList.h>
#include <OpenMS/DATASTRUCTURES/String.h>

#include <utility>

namespace OpenMS
{
  /**
    @brief Hidden Markov Model State class for the Hidden Markov Model
  */
  class OPENMS_DLLAPI HMMState
  {
public:

    /** @name Constructors and destructors
    */
    //@{
    /// default constructor
    HMMState();

    /// copy constructor
    HMMState(const HMMState & state);

    /// constructor with name and visibility option
    HMMState(const String & name, bool hidden = true);

    /// destructor
    virtual ~HMMState();
    //@}

    ///
    HMMState & operator=(const HMMState &);

    /** Accessors
    */
    //@{
    /// sets the name of the state
    void setName(const String & name);

    /// returns the name of the state
    const String & getName() const;

    /// sets the hidden property to the state
    void setHidden(bool hidden);

    /// returns true if the state is hidden
    bool isHidden() const;

    /// adds the given predecessor state to the list
    void addPredecessorState(HMMState * state);

    /// deletes the given predecessor state from the list
    void deletePredecessorState(HMMState * state);

    /// add the given successor state to the list
    void addSuccessorState(HMMState * state);

    /// deletes the given successor state from the list
    void deleteSuccessorState(HMMState * state);

    /// returns the predecessor states of the state
    const std::set<HMMState *> & getPredecessorStates() const;

    /// return the successor states of the state
    const std::set<HMMState *> & getSuccessorStates() const;
    //@}

protected:

    ///
    bool hidden_;

    ///
    String name_;

    ///
    std::set<HMMState *> pre_states_;

    ///
    std::set<HMMState *> succ_states_;
  };


  /**
    @brief  Hidden Markov Model implementation of PILIS

                      Hidden Markov Model implementation suitable for forward conncected HMMs.
                      The HMM is mostly used within PILIS. For further details have a look at
                      the docs of PILIS.
  */
  class OPENMS_DLLAPI HiddenMarkovModel
  {
public:

    /** @name Constructors and destructors
     */
    //@{
    /// default constructor
    HiddenMarkovModel();

    /// copy constructor
    HiddenMarkovModel(const HiddenMarkovModel & hmm_new);

    /// destructor
    virtual ~HiddenMarkovModel();
    //@}

    /// assignment operator
    HiddenMarkovModel & operator=(const HiddenMarkovModel &);

    /** Accessors
    */
    //@{
    /** @brief writes the HMM into a file in GraphML format

            A detailed description of the GraphML format can be found under
            http://graphml.graphdrawing.org/
    */
    void writeGraphMLFile(const String & filename);

    /// writes the HMM into an outstream
    void write(std::ostream & out) const;

    /// returns the transition probability of the given state names
    DoubleReal getTransitionProbability(const String & s1, const String & s2) const;

    /// sets the transition probability of the given state names to prob
    void setTransitionProbability(const String & s1, const String & s2, DoubleReal prob);

    /// return the number of states
    Size getNumberOfStates() const;

    /// registers a new state to the HMM
    void addNewState(HMMState * state);

    /// registers a new state to the HMM
    void addNewState(const String & name);

    /// add a new synonym transition to the given state names
    void addSynonymTransition(const String & name1, const String & name2, const String & synonym1, const String & synonym2);

    /// evaluate the HMM, estimates the transition probabilities from the training
    void evaluate();

    /// trains the HMM; initial probabilities and emission probabilities of the emitting states should be set
    void train();

    /// sets the initial transition probability of the given state to prob
    void setInitialTransitionProbability(const String & state, DoubleReal prob);

    /// clears the initial probabilities
    void clearInitialTransitionProbabilities();

    /// sets the emission probability of the given state to prob
    void setTrainingEmissionProbability(const String & state, DoubleReal prob);

    /// clear the emission probabilities
    void clearTrainingEmissionProbabilities();

    /// enables a transition; adds s1 to the predecessor list of s2 and s2 to the successor list of s1
    void enableTransition(const String & s1, const String & s2);

    /// disables the transition; deletes the nodes from the predeccessor/successor list repsectively
    void disableTransition(const String & s1, const String & s2);

    /// disables all transitions
    void disableTransitions();

    /// calculates the emission probabilities of the HMM (of course only of the non-hidden states)
    void calculateEmissionProbabilities(Map<HMMState *, DoubleReal> & emission_probs);

    /// writes some stats to cerr
    void dump();

    /// writes some info of the forward "matrix" to cerr
    void forwardDump();

    /// builds a synonyms structure, needed when synonyms are used
    //void buildSynonyms();

    /// estimates the transition probabilities of not trained transitions by averages similar trained ones
    void estimateUntrainedTransitions();

    /// returns the state with the given name
    HMMState * getState(const String & name);

    /// returns the state with the given name
    const HMMState * getState(const String & name) const;

    /// clears all data
    void clear();

    /// sets the pseudo count that are added instead of zero
    void setPseudoCounts(DoubleReal pseudo_counts);

    /// returns the pseudo counts
    DoubleReal getPseudoCounts() const;

    void setVariableModifications(const StringList & modifications);
    //@}

protected:

    /// disables the transition; deletes the nodes from the predeccessor/successor list repsectively
    void disableTransition_(HMMState * s1, HMMState * s2);

    /// enables a transition; adds s1 to the predecessor list of s2 and s2 to the successor list of s1
    void enableTransition_(HMMState * s1, HMMState * s2);

    /// sets the emission probability of the given state to prob
    void setTrainingEmissionProbability_(HMMState * state, DoubleReal prob);

    /// sets the transition probability of the given states to prob
    void setTransitionProbability_(HMMState * s1, HMMState * s2, DoubleReal prob);

    /// returns the transition probability of the given states
    DoubleReal getTransitionProbability_(HMMState * s1, HMMState * s2) const;


    /// performs the forward algorithm
    void calculateForwardPart_();

    /// performs the backward algorithm
    void calculateBackwardPart_();

    /// returns the forward variable
    DoubleReal getForwardVariable_(HMMState *);

    /// returns the backward variable
    DoubleReal getBackwardVariable_(HMMState *);

private:

    // transition probs
    Map<HMMState *, Map<HMMState *, DoubleReal> > trans_;

    // transition prob counts
    Map<HMMState *, Map<HMMState *, DoubleReal> > count_trans_;

    Map<HMMState *, Map<HMMState *, std::vector<DoubleReal> > > count_trans_all_;

    // all transition probs of all training steps (for model checking)
    Map<HMMState *, Map<HMMState *, std::vector<DoubleReal> > > train_count_trans_all_;

    // number of training steps of the transitions
    Map<HMMState *, Map<HMMState *, Size> > training_steps_count_;

    // forward variables
    Map<HMMState *, DoubleReal> forward_;

    // backward variables
    Map<HMMState *, DoubleReal> backward_;

    // name to state Mapping
    Map<String, HMMState *> name_to_state_;

    // emission probabilities
    Map<HMMState *, DoubleReal> train_emission_prob_;

    // initial transition probabilities
    Map<HMMState *, DoubleReal> init_prob_;

    // all states of the HMM
    std::set<HMMState *> states_;

    // trained transitions
    std::set<std::pair<HMMState *, HMMState *> > trained_trans_;

    // synonym transitions Mapping
    Map<String, Map<String, std::pair<String, String> > > synonym_trans_names_;

    // synonym transitions
    Map<HMMState *, Map<HMMState *, std::pair<HMMState *, HMMState *> > > synonym_trans_;

    // transitions which are enabled
    Map<HMMState *, std::set<HMMState *> > enabled_trans_;

    // pseudocounts used in this instance
    DoubleReal pseudo_counts_;

    // copy all the stuff from one HMM to this
    void copy_(const HiddenMarkovModel & source);

    StringList var_modifications_;
  };
}
#endif