Interface Distribution

All Superinterfaces:
Changeable
All Known Subinterfaces:
OrderNDistribution
All Known Implementing Classes:
AbstractDistribution, AbstractOrderNDistribution, GapDistribution, PairDistribution, SimpleDistribution, TranslatedDistribution, UniformDistribution, UntrainableDistribution

public interface Distribution extends Changeable

An encapsulation of a probability distribution over the Symbols within an alphabet.

A distribution can be implemented as a map from symbol to probability. It is more correct to think of them as being integrals or sums over probability dencity funcitons. In this world view, getWeight should look at the getMatches of the symbol it is given and then perform the apropreate sum or integral to return the probability of something within that set of symbols being emitted.

This interface should handle the case of emitting an ambiguity symbol. This should be just the sum of the probabiltiy of emitting each matching symbol. It is up to the code using the Distribution instance to divide out the null model appropreately.

Since:
1.0
Author:
Matthew Pocock
  • Field Details

    • WEIGHTS

      static final ChangeType WEIGHTS

      Whenever a distribution changes the values that would be returned by getWeight, they should fire a ChangeEvent with this object as the type.

      If the whole distribution changes, then the change and previous fields of the ChangeEvent should be left null. If only a single weight is modified, then change should be of the form Object[] { symbol, new Double(newVal) } and previous should be of the form Object[] { symbol, new Double(oldVal) }

    • NULL_MODEL

      static final ChangeType NULL_MODEL

      Whenever the null model distribution changes the values that would be returned by getWeight, either by being edited or by being replaced, a ChangeEvent with this object as the type should be thrown.

      If the null model has changed its weights, then the ChangeEvent should refer back to the ChangeEvent from the null model.

  • Method Details

    • getAlphabet

      The alphabet from which this spectrum emits symbols.
      Returns:
      the Alphabet associated with this spectrum
    • getWeight

      Return the probability that Symbol s is emitted by this spectrum.

      If the symbol is ambiguou, then it is the sum of the probability that each one of the matching symbols was emitted.

      Parameters:
      s - the Symbol emitted
      Returns:
      the probability of emitting that symbol
      Throws:
      IllegalSymbolException - if s is not from this state's alphabet
    • setWeight

      Set the probability or odds that Symbol s is emitted by this state.
      Parameters:
      s - the Symbol emitted
      w - the probability of emitting that symbol
      Throws:
      IllegalSymbolException - if s is not from this state's alphabet, or if it is an ambiguity symbol and the implementation can't handle this case
      ChangeVetoException - if this state does not allow weights to be tampered with, or if one of the listeners vetoed this change
    • sampleSymbol

      Sample a symbol from this state's probability distribution.
      Returns:
      the symbol sampled
    • getNullModel

      Retrieve the null model Distribution that this Distribution recognizes.
      Returns:
      the apropriate null model
    • setNullModel

      Set the null model Distribution that this Distribution recognizes.
      Parameters:
      nullDist - the new null model Distribution
      Throws:
      IllegalAlphabetException - if the null model has the wrong alphabet
      ChangeVetoException - if this Distirbution doesn't support setting the null model, or if one of its listeners objects
    • registerWithTrainer

      Register this distribution with a training context.

      This should be invoked from within dtc.addDistribution(). This method is responsible for constructing a suitable DistributionTrainer instance and registering it by calling dtc.registerDistributionTrainer(this, trainer). If the distribution is a view onto another distribution, it can force the other to be registered by calling dtc.addDistribution(other), and can then get on with registering it's own trainer.

      Parameters:
      dtc - the DistributionTrainerContext with witch to register a trainer