MyReallyAbstractGainAlgo Class Reference

A super-duper rilly, rilly abstract class for function prediction algorithms in GAIN. This class only contains basic methods that all algorithms need. More...

#include <gain-algorithm.h>

Inheritance diagram for MyReallyAbstractGainAlgo:

List of all members.

Public Member Functions
virtual void	initialiseAlgorithm ()
virtual string	getName ()=0
virtual void	pipeline (const BioFunction &function)=0
virtual void	printStatistics (ostream &statsStream)

---

Detailed Description

A super-duper rilly, rilly abstract class for function prediction algorithms in GAIN. This class only contains basic methods that all algorithms need.

Implementing a New Algorithm in GAIN.

GAIN contains an easy-to-use framework to implement new function prediction algorithms. Here are the steps that you will need to take:

1. Decide if the algorithm uses both positive and negative examples or only positive examples. We call algorithms that use both positive and negative example "one-versus-all" algorithms and those that use only positive examples "one-versus-none" algorithms. The steps below show how to add a one-versus-none algorithm called "local" to the GAIN code. The steps for adding a one-versus-all algorithm are analogous.

2. Give your algorithm a short mnemonic name, e.g., "local".

3. Add your algorithm's name as a possible option to the one-versus-none and the shorter ovn option in gain.ggo. This change will allow the user to invoke the algorithm with the --one-versus-none local or the --ovn local option.

4. Create a class called MyOneVersusNoneLocalGainAlgo as a subclass of MyOneVersusNoneGainAlgo in gain-algorithm.h. Document the class, especially by summarising the algorithm. Override any methods in MyOneVersusNoneGainAlgo that you need to.

   • You will have to override MyOneVersusNoneGainAlgo::getName(). In this case, MyOneVersusNoneLocalGainAlgo::getName() returns "OneVersusNoneLocal".

   • You will also have to override MyOneVersusNoneGainAlgo::computeState(), since this method is at the very core of every algorithm: it specifies how the state of a node is to be computed, usually by combining the states of the node's neighbours.

   • You may also have to override MyOneVersusNoneGainAlgo::run(), for instance if you do something extra after computing all node states (see

     See also:

     MyOneVersusNoneFunctionalFlowGainAlgo::run()) or if you have to control the number of times to loop over the nodes (see

     MyOneVersusNoneLocalGainAlgo::run()).

5. Add code to the function addAlgorithms in gain.C that will create an instance of MyOneVersusNoneLocalGainAlgo().

---

Member Function Documentation

virtual string MyReallyAbstractGainAlgo::getName

(

)

[pure virtual]

Return the name of the algorithm.

Implemented in MyOneVersusAllGeneManiaGainAlgo, MyOneVersusNoneLocalGainAlgo, MyOneVersusNoneFunctionalFlowGainAlgo, MyOneVersusNoneSinkSourceGainAlgo, MyOneVersusAllSinkSourceGainAlgo, MyOneVersusAllSVMLightTransductiveGainAlgo, MyOneVersusAllSVMLightGainAlgo, MyOneVersusAllLibSVMGainAlgo, MyOneVersusAllHierarchicalHopfieldGainAlgo, MyOneVersusAllSemiHierarchicalHopfieldGainAlgo, MyOneVersusAllMincutGainAlgo, MyOneVersusAllHopfieldGainAlgo, MyOneVersusAllLocalGainAlgo, MyOneVersusNoneGainAlgo, MyOneVersusAllGainAlgo, MyAbstractGainAlgo< StateType >, and MyAbstractGainAlgo< MyGainTriStateType >.

virtual void MyReallyAbstractGainAlgo::initialiseAlgorithm

(

)

[inline, virtual]

Initialise internal data structures for the algorithm.

Reimplemented in MyOneVersusAllSemiHierarchicalHopfieldGainAlgo.

virtual void MyReallyAbstractGainAlgo::pipeline

(

const BioFunction &

function

)

[pure virtual]

Run the complete pipeline for the algorithm.

Implemented in MyAbstractGainAlgo< StateType >, and MyAbstractGainAlgo< MyGainTriStateType >.

virtual void MyReallyAbstractGainAlgo::printStatistics

(

ostream &

statsStream

)

[inline, virtual]

Print statistics about the performance of the algorithm, after processing all functions.

Reimplemented in MyOneVersusAllSemiHierarchicalHopfieldGainAlgo, and MyOneVersusAllGainAlgo.

---

The documentation for this class was generated from the following file:

• /home/poirel/src/c++/biorithm/gain/gain-algorithm.h