com.chemaxon.descriptors.common.unguarded

Interface UnguardedDissimilarityCalculator<T>

Type Parameters:

T - Type of compared objects

All Superinterfaces:

Serializable

All Known Implementing Classes:

BinaryLongCommonpart, BinaryLongCommonpartNormalized, BinaryLongDice, BinaryLongEuclidean, BinaryLongEuclideanNormalized, BinaryLongManhattan, BinaryLongManhattanNormalized, BinaryLongPetke, BinaryLongSimpson, BinaryLongTanimoto, BinaryLongTversky, FloatsEuclid, FloatsEuclidSqr, FloatsManhattan, FloatsMaxDiff, FloatsTanimoto, FloatsTversky, ScaledBytesFvEuclid, ScaledBytesFvTanimoto, ScaledShortsFvEuclid, SerializingUnguardedDissimilarityCalculator, SlowUnguardedComparator

@Beta
public interface UnguardedDissimilarityCalculator<T>
extends Serializable

Unguarded dissimilarity calculator. Unguarded calculators are used in performance critical environment. Parts of the descriptor API constrains are enforced on the higher usage level. These further API constraints are not necessarily checked on this level, so extreme care is needed for usage.

Calculated dissimilarity is represented as raw and normalized value. Raw values

• Are ordered, smaller (possibly negative) numerical value means more similar according to the represented metric
• Are possibly scaled using a non linear bijective function

Expected contracts for implementations:

• Instances are immutable.
• Instances are thread safe.
• Instances are Serializable
• Instances might have construction time parameter(s), like fingerprint length

Relaxed API constraints which are expected to be enforced on higher level:

• Descriptor generator compatibility checking (see Guarded.getGuardObject()
• Preconditions of parameters (check for null, lengths, etc)
• Validity of parameters (values, ranges)

Method Summary

Modifier and Type	Method and Description
double	denormalize(double dissimilarity) Calculate raw value for a normalized dissimilarity value.
double	dissimilarity(T target, T query) Compare dissimilarity of two objects.
double	normalize(double rawValue) Normalize dissimilarity value returned by dissimilarity(java.lang.Object, java.lang.Object).

Method Detail

dissimilarity

double dissimilarity(T target,
                     T query)

Compare dissimilarity of two objects. Raw dissimilarity score is returned which

Note that implementations might not check even basic consistency of compared objects.

Implementations must return valid real numbers which are smaller than Double.MAX_VALUE.

Parameters:

target - Target to compare

query - Query to compare

Returns:

Raw score. Lower score represents more similar object

normalize

double normalize(double rawValue)

Normalize dissimilarity value returned by dissimilarity(java.lang.Object, java.lang.Object).

Parameters:

rawValue - Raw dissimilarity value

Returns:

Normalized value

denormalize

double denormalize(double dissimilarity)

Calculate raw value for a normalized dissimilarity value. This function can be used where a cutoff defined in terms of dissimilarity is implemented.

Parameters:

dissimilarity - Normalized dissimilarity value

Returns:

Raw value