M
- hypothesis model type (upper bound)I
- input symbol typeD
- output typepublic interface LearnerVariantList<M,I,D>
Usually, there should be one integration test class per learning algorithm. However, in many cases a single learning algorithm can be configured in numerous ways, all (or many) of which should be tested independently. Due to the large number of possible combinations, it is undesirable to create a single integration test class for each configuration; instead, these variants should be configured and created programmatically. The purpose of the variant list is to offer a convenient interface for storing all these variants.
Modifier and Type | Interface and Description |
---|---|
static interface |
LearnerVariantList.DFALearnerVariantList<I> |
static interface |
LearnerVariantList.MealyLearnerVariantList<I,O> |
static interface |
LearnerVariantList.MealySymLearnerVariantList<I,O> |
static interface |
LearnerVariantList.OneSEVPALearnerVariantList<I> |
Modifier and Type | Method and Description |
---|---|
void |
addLearnerVariant(String name,
LearningAlgorithm<? extends M,I,D> learner)
Adds a learner variant with the default maximum number of rounds (i.e., the size of the target automaton) to the
list.
|
void |
addLearnerVariant(String name,
LearningAlgorithm<? extends M,I,D> learner,
int maxRounds)
Adds a learner variant with a given maximum number of rounds to the list.
|
void addLearnerVariant(String name, LearningAlgorithm<? extends M,I,D> learner)
This is a convenience method, equivalent to invoking addLearnerVariant(name, learner, -1)
.
name
- the name of the variantlearner
- the algorithm instance for this variantvoid addLearnerVariant(String name, LearningAlgorithm<? extends M,I,D> learner, int maxRounds)
name
- the name of the variantlearner
- the algorithm instance for this variantmaxRounds
- the maximum number of rounds for the specified target automaton. If a value less than or equal to zero is
specified, the default maximum number of rounds (the size of the target automaton) is assumed.Copyright © 2018. All rights reserved.