CompositionSetDistanceFilter (magpie API)

java.lang.Object
- magpie.data.utilities.filters.BaseDatasetFilter
- - magpie.data.utilities.filters.CompositionSetDistanceFilter

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, Options
```
public class CompositionSetDistanceFilter
extends BaseDatasetFilter
implements java.lang.Cloneable
```
Filter based on whether the distance of a composition from any other composition in a dataset is past a threshold. The distance between two compositions is based on the distance between vectors defined by element fractions. We use either the Manhattan distance (equivalent to the L₁ norm used in CompositionDistanceFilter) or the Euclidean distance (L₂ norm). The distance between a composition and the set is defined as the minimum distance between a composition and any entry in the dataset.
Entries are labeled based on whether they are farther than a certain threshold.

Note: Euclidean distance is much faster than Manhattan. Weka's algorithms for fast neighbor search are not designed to work with Manhattan distance.

Developer note: This class uses different distance metrics than CompositionDistanceFilter< in order to use the NearestNeighbourSearch classes from Weka to accelerate finding the closest compositions.

Usage: $<dataset> <-mahanttan|-euclidean> <threshold>
dataset: CompositionDataset containing all compositions being considered
-manhattan: Use Manhattan distance metric
-euclidean: Use Euclidean distance metric
threshold: Distance threshold

See Also:

Serialized Form

Constructor Summary

Constructors
Constructor and Description

CompositionSetDistanceFilter()

Constructors
Constructor and Description
`CompositionSetDistanceFilter()`

Method Summary

All Methods Static Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`void`	`addComposition(CompositionEntry entry)` Add a new composition to the dataset.
`void`	`addCompositions(java.util.Collection<CompositionEntry> comps)` Add a list of compositions to the set
`void`	`clearCompositions()` Clear the list of compositions in set
`CompositionSetDistanceFilter`	`clone()`
`double`	`computeDistance(CompositionEntry entry)` Compute the distance between a composition and the set of other compositions stored in this object.
`protected static weka.core.Instances`	`convertCompositionsToInstances(java.util.Collection<CompositionEntry> entries)` Convert a collection of entries to a Weka Instances object.
`static weka.core.Instance`	`convertCompositionToInstance(CompositionEntry entry)` Convert a composition to a Weka instance.
`boolean[]`	`label(Dataset D)` Given a dataset, determine which entries passes the filter.
`void`	`makeNeighborSearchTool()` Generate a tool to enable fast searching for nearest compositions.
`protected int`	`parallelMinimum()` Minimum number of entries to label in parallel.
`java.lang.String`	`printUsage()` Print out required format for options.
`void`	`setCompositions(CompositionDataset data)` Set the list of compositions to be considered
`void`	`setDistanceThreshold(double dist)` Set the distance threshold
`void`	`setOptions(java.util.List<java.lang.Object> Options)` Set any options for this object.
`void`	`setUseEuclidean()` Use Euclidean distance as the distance metric
`void`	`setUseManhattan()` Use Manhattan distance as the distance metric
`void`	`setUseManhattan(boolean manhattan)` Set whether to use Manhattan (vs Euclidean) distance
`void`	`train(Dataset TrainingSet)` Train a dataset splitter, if necessary

Methods inherited from class magpie.data.utilities.filters.BaseDatasetFilter
filter, parallelLabel, setExclude, toExclude

Methods inherited from class java.lang.Object
equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Constructor Detail
  - CompositionSetDistanceFilter
```
public CompositionSetDistanceFilter()
```
- Method Detail
  - setOptions
```
public void setOptions(java.util.List<java.lang.Object> Options)
                throws java.lang.Exception
```
    Description copied from interface: Options
    
    Set any options for this object. If parsing the options fails, return an Exception with the usage as its message.
    
    Specified by:
    
    setOptions in interface Options
    
    Parameters:
    
    Options - Array of options as Objects - can be null
    
    Throws:
    
    java.lang.Exception - if problem with inputs
  - clone
```
public CompositionSetDistanceFilter clone()
```
    Overrides:
    
    clone in class java.lang.Object
  - printUsage
```
public java.lang.String printUsage()
```
    Description copied from interface: Options
    
    Print out required format for options. For consistency, should start with "Usage: ".
    
    Specified by:
    
    printUsage in interface Options
    
    Returns:
    
    Usage for this function
  - setUseManhattan
```
public void setUseManhattan(boolean manhattan)
```
    Set whether to use Manhattan (vs Euclidean) distance
    
    Parameters:
    
    manhattan - Desired setting
  - setUseManhattan
```
public void setUseManhattan()
```
    Use Manhattan distance as the distance metric
    
    See Also:
    
    setUseEuclidean()
  - setUseEuclidean
```
public void setUseEuclidean()
```
    Use Euclidean distance as the distance metric
    
    See Also:
    
    setUseManhattan()
  - clearCompositions
```
public void clearCompositions()
```
    Clear the list of compositions in set
  - addComposition
```
public void addComposition(CompositionEntry entry)
```
    Add a new composition to the dataset.
    
    Parameters:
    
    entry - Entry to be added
  - addCompositions
```
public void addCompositions(java.util.Collection<CompositionEntry> comps)
```
    Add a list of compositions to the set
    
    Parameters:
    
    comps - Collection of compositions to be added
  - setCompositions
```
public void setCompositions(CompositionDataset data)
```
    Set the list of compositions to be considered
    
    Parameters:
    
    data - Dataset containing compositions to use as dataset
  - setDistanceThreshold
```
public void setDistanceThreshold(double dist)
                          throws java.lang.Exception
```
    Set the distance threshold
    
    Parameters:
    
    dist - Desired distance threshold
    
    Throws:
    
    java.lang.Exception - If distance is ≤ 0
  - parallelMinimum
```
protected int parallelMinimum()
```
    Description copied from class: BaseDatasetFilter
    
    Minimum number of entries to label in parallel.
    
    Overrides:
    
    parallelMinimum in class BaseDatasetFilter
    
    Returns:
    
    Implementer-specified minimum threshold
  - train
```
public void train(Dataset TrainingSet)
```
    Description copied from class: BaseDatasetFilter
    
    Train a dataset splitter, if necessary
    
    Specified by:
    
    train in class BaseDatasetFilter
    
    Parameters:
    
    TrainingSet - Dataset to use for training
  - label
```
public boolean[] label(Dataset D)
```
    Description copied from class: BaseDatasetFilter
    
    Given a dataset, determine which entries passes the filter. Please ensure that D remains unaltered (clone if you must)
    
    Specified by:
    
    label in class BaseDatasetFilter
    
    Parameters:
    
    D - Dataset to be labeled
    
    Returns:
    
    A list of booleans marking whether an entry passes
  - convertCompositionToInstance
```
public static weka.core.Instance convertCompositionToInstance(CompositionEntry entry)
```
    Convert a composition to a Weka instance. Puts it in a form that can be used with Weka's neighbor search methods. Attributes for the instance are the fraction of each element.
    
    Parameters:
    
    entry - Entry to be converted
    
    Returns:
    
    Instance describing the composition of the entry
  - convertCompositionsToInstances
```
protected static weka.core.Instances convertCompositionsToInstances(java.util.Collection<CompositionEntry> entries)
```
    Convert a collection of entries to a Weka Instances object. Another step in getting ready to make a Weka nearest neighbor search tool.
    
    Parameters:
    
    entries - Entries to be converted
    
    Returns:
    
    Weka dataset object
  - makeNeighborSearchTool
```
public void makeNeighborSearchTool()
```
    Generate a tool to enable fast searching for nearest compositions. Stored in the class.
  - computeDistance
```
public double computeDistance(CompositionEntry entry)
```
    Compute the distance between a composition and the set of other compositions stored in this object. Distance is defined by your choice of setUseManhattan()
    
    Parameters:
    
    entry - Composition of entry
    
    Returns:
    
    Distance from set

Class CompositionSetDistanceFilter

Constructor Summary

Method Summary

Methods inherited from class magpie.data.utilities.filters.BaseDatasetFilter

Methods inherited from class java.lang.Object

Constructor Detail

CompositionSetDistanceFilter

Method Detail

setOptions

clone

printUsage

setUseManhattan

setUseManhattan

setUseEuclidean

clearCompositions

addComposition

addCompositions

setCompositions

setDistanceThreshold

parallelMinimum

train

label

convertCompositionToInstance

convertCompositionsToInstances

makeNeighborSearchTool

computeDistance