Package org.anchoranalysis.feature.bean

Class Feature<T extends FeatureInput>

Object
AnchorBean<Feature<T>>
InitializableBean<Feature<T>,FeatureInitialization>
Feature<T>
Type Parameters:
T - input-type from which a measurement is calculated.
Direct Known Subclasses:
FeatureAllMemo, FeatureBinary, FeatureEnergy, FeatureFromList, FeatureGeneric, FeatureHistogram, FeatureImageMetadata, FeatureObjectCollection, FeaturePairMemo, FeaturePairObjects, FeatureResults, FeatureSingleMemo, FeatureSingleObject, FeatureStack, FeatureUnary, FromDimensionsBase
public abstract class Feature<T extends FeatureInput> extends InitializableBean<Feature<T>,FeatureInitialization>
Base class for a measurement that calculates a result (double) given input parameters.

This is a key base class in the Anchor framework, and many algorithms and procedures use a Feature on particular input-types to calculate measurements in a flexible way, and often to use these measurements in a machine-learning algorithm.

See Feature (Machine learning) on Wikipedia for general background on the concept of a feature in Machine Learning.

Author:
Owen Feehan

  • Constructor Details

    • Feature

      protected Feature()
      Creates with the default initializer.

    • Feature

      protected Feature(BeanInitializer<FeatureInitialization> propertyInitializer)
      Creates with a custom initializer.
      Parameters:
      propertyInitializer - the custom initializer.

  • Method Details

    • onInitialization

      public void onInitialization(FeatureInitialization initialization) throws InitializeException
      Called after initialization.

      An empty implementation is provided, to be overridden as needed in the sub-classes.

      Overrides:
      onInitialization in class InitializableBean<Feature<T extends FeatureInput>,FeatureInitialization>
      Parameters:
      initialization - parameters used for initialization.
      Throws:
      InitializeException - if initialization does not successfully complete.

    • inputType

      public abstract Class<? extends FeatureInput> inputType()
      The class corresponding to feature input-type.

      i.e. corresponding to the T template parameter.

      Returns:
      the class.

    • describeBean

      public final String describeBean()
      Description copied from class: AnchorBean
      A (maybe longer) description identifying the bean and perhaps its key parameters.

      By default, it returns the same as AnchorBean.getBeanName() but beans can optionally override it

      Overrides:
      describeBean in class AnchorBean<Feature<T extends FeatureInput>>
      Returns:
      either the short-name of the bean, or a longer description

    • descriptionLong

      public String descriptionLong()
      A long human-readable description of the feature and some or all of its parameterization.

      This can be overwritten to create an alternative description for features. By default string returned by describeBean() is used.

      Returns:
      the description.

    • duplicateChangeName

      public Feature<T> duplicateChangeName(String customName)
      Duplicates the feature as per AnchorBean.duplicateBean() but sets a particular custom-name.
      Parameters:
      customName - the custom-name to set.
      Returns:
      a duplicated (deep copy of bean attributes) feature, identical to current feature, but with the specified custom-name.

    • getFriendlyName

      public String getFriendlyName()
      A user-friendly human-readable name for the Feature.

      If a custom-name has been assigned, this is returned, otherwise the descriptionLong().

      Returns:
      the user-friendly human-readable name.

    • createListChildFeatures

      public final FeatureList<FeatureInput> createListChildFeatures() throws BeanMisconfiguredException
      Returns a list of Features that exist as bean-properties of this feature, either directly or in lists.

      It does not recurse.

      It ignores features that are referenced from elsewhere.

      Returns:
      the list of features.
      Throws:
      BeanMisconfiguredException - if the feature-beans are not structured as expected.

    • describeParameters

      public String describeParameters()
      A human-readable description of the parameterization of the bean.
      Returns:
      the description.

    • toString

      public String toString()
      Description copied from class: AnchorBean
      By default, we use AnchorBean.describeBean() as the string representation of the bean.
      Overrides:
      toString in class AnchorBean<Feature<T extends FeatureInput>>
      Returns:
      the string representation

    • castAs

      public <S extends T> Feature<S> castAs()
      Casts the feature to having a different input-type.

      Note that no active compile-type check occurs, so be careful that this is used appropriately.

      Type Parameters:
      S - the type to cast to, which should be a sub-type of the existing-type.
      Returns:
      the same instance, cast as having a different input-type.

    • calculateCheckInitialized

      public double calculateCheckInitialized(FeatureCalculationInput<T> input) throws FeatureCalculationException
      Calculates the result of a feature and throws an exception if the feature has not been initialized.

      It is not recommended to directly use this method to calculate a feature's value, but rather to:

      • Externally, for code outside a Feature's implementation, please use the feature-session classes, who will perform the calculation using caching to reduce redundant computation.
      • Internally, for code inside a Feature's implementation, please call FeatureCalculationInput.calculate(Feature) and similar methods, who will back-reference the session above, to take advantage of the caching.
      Parameters:
      input - the input to the calculation.
      Returns:
      the feature-value corresponding to input for this feature.
      Throws:
      FeatureCalculationException - if the feature has not been initialized.

    • beforeCalc

      protected void beforeCalc(FeatureInitialization initialization) throws InitializeException
      Dummy method, that children can optionally override.
      Parameters:
      initialization - initialization parameters.
      Throws:
      InitializeException - if initialization cannot complete successfully.

    • calculate

      protected abstract double calculate(FeatureCalculationInput<T> input) throws FeatureCalculationException
      Calculates a value for some input.
      Parameters:
      input - the input to the calculation.
      Returns:
      the result of the calculation.
      Throws:
      FeatureCalculationException - if the calculation cannot successfully complete.

    • assignTo

      protected void assignTo(Feature<FeatureInput> target)
      Copies fields in this (base) class to target.

      This is intended to be called by sub-classes to help when duplicating.

      Parameters:
      target - the object fields are assigned to.

    • getCustomName

      public String getCustomName()
      An optional additional name that be associated with the feature (defaults to an empty string).

    • setCustomName

      public void setCustomName(String customName)
      An optional additional name that be associated with the feature (defaults to an empty string).