autoweka.instancegenerators
Class MultiLevel
java.lang.Object
autoweka.InstanceGenerator
autoweka.instancegenerators.RandomSubSampling
autoweka.instancegenerators.MultiLevel
public class MultiLevel
- extends RandomSubSampling
Experimental InstanceGenerator that takes as input a child classifier, and creates multiple levels of training data.
For example, when used with the CV Generator, this will produce instances that do 10-CV over 30% of the data, then instance that do 10-CV on 60%, then 10-CV on all of the data
instanceGeneratorArguments:
This is a method that requires a child, instance generator, so the format is ARGS[$]CHILDCLASSNAME[$]CHILDARGS
levelSeed - the seed used to split up into levels
levelPercent - the percentage of data to use at each level
levelBias - the bias towards a uniform partition
numLevels - the total number of levels to use
instance string format:
levelSeed - the seed used to split up into levels
levelPercent - the percentage of data to use at each level
levelBias - the bias towards a uniform partition
level - the current level before passing off to the child
Method Summary |
Instances |
_getTestingFromParams(java.lang.String params)
Subclass implementation for getting the testing data given the param string |
Instances |
_getTrainingFromParams(java.lang.String params)
Subclass implementation for getting the training data given the param string |
java.util.Map<java.lang.String,java.util.Map<java.lang.String,java.lang.String>> |
getAllInstanceFeatures(java.lang.String params)
Some SMBO methods can leverage extra features about an Instance, this method should return a map of feature/value pairs for all instances |
java.util.List<java.lang.String> |
getAllInstanceStrings(java.lang.String params)
Gets a list of all the 'params' Strings that can be used with this InstanceGenerator. |
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
MultiLevel
public MultiLevel(java.lang.String instanceFileName)
MultiLevel
public MultiLevel(InstanceGenerator generator)
MultiLevel
public MultiLevel(Instances training,
Instances testing)
_getTrainingFromParams
public Instances _getTrainingFromParams(java.lang.String params)
- Description copied from class:
InstanceGenerator
- Subclass implementation for getting the training data given the param string
- Overrides:
_getTrainingFromParams
in class RandomSubSampling
_getTestingFromParams
public Instances _getTestingFromParams(java.lang.String params)
- Description copied from class:
InstanceGenerator
- Subclass implementation for getting the testing data given the param string
- Overrides:
_getTestingFromParams
in class RandomSubSampling
getAllInstanceStrings
public java.util.List<java.lang.String> getAllInstanceStrings(java.lang.String params)
- Description copied from class:
InstanceGenerator
- Gets a list of all the 'params' Strings that can be used with this InstanceGenerator.
For example, the for 10-fold Cross Validation this method would return 10 Strings defining each fold
- Overrides:
getAllInstanceStrings
in class RandomSubSampling
getAllInstanceFeatures
public java.util.Map<java.lang.String,java.util.Map<java.lang.String,java.lang.String>> getAllInstanceFeatures(java.lang.String params)
- Description copied from class:
InstanceGenerator
- Some SMBO methods can leverage extra features about an Instance, this method should return a map of feature/value pairs for all instances
- Overrides:
getAllInstanceFeatures
in class InstanceGenerator