BilinearLearnerParameters (OpenIMAJ master project 1.3.10 API)

java.lang.Object
- java.util.AbstractMap<K,V>
- - java.util.HashMap<String,Object>
  - - org.openimaj.ml.linear.learner.LearningParameters
    - - org.openimaj.ml.linear.learner.BilinearLearnerParameters

All Implemented Interfaces:

Serializable, Cloneable, Map<String,Object>, WriteableASCII
```
public class BilinearLearnerParameters
extends LearningParameters
implements WriteableASCII
```
Parameters used to control a BilinearSparseOnlineLearner

Author:

Sina Samangooei (ss@ecs.soton.ac.uk)

See Also:

Serialized Form

Nested Class Summary
- Nested classes/interfaces inherited from class org.openimaj.ml.linear.learner.LearningParameters
  LearningParameters.Placeholder
- Nested classes/interfaces inherited from class java.util.AbstractMap
  AbstractMap.SimpleEntry<K,V>, AbstractMap.SimpleImmutableEntry<K,V>
- Nested classes/interfaces inherited from interface java.util.Map
  Map.Entry<K,V>

Field Summary

Fields
Modifier and Type	Field and Description
`static String`	`BIAS` whether a bias component is added to w and u.
`static String`	`BIASINITSTRAT` The initialisation strategy for BIAS.
`static String`	`BICONVEX_MAXITER` The maximum number of iterations in the biconvex iterative stage.
`static String`	`BICONVEX_TOL` The threshold of the ratio between the (sum(new_w - old_w) + sum(new_u - old_u)) / (sum(old_u) + sum(old_w)) i.e.
`static String`	`DAMPENING` The value of w, u and beta are updated each time data is added s.t.
`static String`	`ETA_GAMMA` The ammount by which ETA is made to increase each iteration
`static String`	`ETA0_BIAS` The weighting of the subgradient of BIAS, weighted down each ETASTEPS number of iterations of the biconvex scheme, defaults to eta0 (0.05)
`static String`	`ETA0_U` The weighting of the subgradient of U, weighted down each ETASTEPS number of iterations of the biconvex scheme, defaults to 0.05
`static String`	`ETA0_W` The weighting of the subgradient of W, weighted down each ETASTEPS number of iterations of the biconvex scheme, defaults to 0.05
`static String`	`ETASTEPS` The steps at which point the eta parameter is reduced, defaults to 3
`static String`	`EXPANDEDUINITSTRAT` The initialisation strategy for U when it is expanded.
`static String`	`EXPANDEDWINITSTRAT` The initialisation strategy for W when it is expanded.
`static String`	`FORCE_SPARCITY` Should all parameter matricies be held `SparseMatrix` instances and therefore remain sparse.
`static String`	`LAMBDA` The parameter of the regulariser for both W and U
`static String`	`LAMBDA_U` The parameter of the regulariser for U, defaults to LAMBDA
`static String`	`LAMBDA_W` The parameter of the regulariser for W, defaults to LAMBDA
`static String`	`LOSS` The loss function, defaults to `SquareMissingLossFunction`
`static String`	`REGUL` The regularisation function, defaults to `L1L2Regulariser`
`static String`	`SEED` The random seed of any randomised components of this learner (usually initialisation).
`static String`	`UINITSTRAT` The initialisation strategy for U.
`static String`	`WINITSTRAT` The initialisation strategy for W.
`static String`	`Z_STANDARDISE` Whether the Vprime and Dprime matrices should be zscore standardised

Constructor Summary

Constructors
Constructor and Description

BilinearLearnerParameters()
sets up the defaults

Constructors
Constructor and Description
`BilinearLearnerParameters()` sets up the defaults

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`String`	`asciiHeader()` Header for ascii output.
`BilinearLearnerParameters`	`clone()`
`String`	`toString()`
`void`	`writeASCII(PrintWriter out)` Write the content of this as ascii to out.

Methods inherited from class org.openimaj.ml.linear.learner.LearningParameters
getTyped

Methods inherited from class java.util.HashMap
clear, compute, computeIfAbsent, computeIfPresent, containsKey, containsValue, entrySet, forEach, get, getOrDefault, isEmpty, keySet, merge, put, putAll, putIfAbsent, remove, remove, replace, replace, replaceAll, size, values

Methods inherited from class java.util.AbstractMap
equals, hashCode

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

Methods inherited from interface java.util.Map
equals, hashCode

- Field Detail
  - BIAS
```
public static final String BIAS
```
    whether a bias component is added to w and u. Default is false.
    
    See Also:
    
    Constant Field Values
  - SEED
```
public static final String SEED
```
    The random seed of any randomised components of this learner (usually initialisation). Defaults to -1 (i.e. no seed)
    
    See Also:
    
    Constant Field Values
  - WINITSTRAT
```
public static final String WINITSTRAT
```
    The initialisation strategy for W. Defaults to a SparseRandomInitStrategy with sparsity set to 0.5
    
    See Also:
    
    Constant Field Values
  - UINITSTRAT
```
public static final String UINITSTRAT
```
    The initialisation strategy for U. Defaults to a SparseRandomInitStrategy with sparsity set to 0.5
    
    See Also:
    
    Constant Field Values
  - EXPANDEDWINITSTRAT
```
public static final String EXPANDEDWINITSTRAT
```
    The initialisation strategy for W when it is expanded. Defaults to the context aware CurrentWMean
    
    See Also:
    
    Constant Field Values
  - EXPANDEDUINITSTRAT
```
public static final String EXPANDEDUINITSTRAT
```
    The initialisation strategy for U when it is expanded. Defaults to the context aware CurrentUMean
    
    See Also:
    
    Constant Field Values
  - BIASINITSTRAT
```
public static final String BIASINITSTRAT
```
    The initialisation strategy for BIAS. Defaults to a SparseZerosInitStrategy
    
    See Also:
    
    Constant Field Values
  - BICONVEX_MAXITER
```
public static final String BICONVEX_MAXITER
```
    The maximum number of iterations in the biconvex iterative stage. defaults to 3
    
    See Also:
    
    Constant Field Values
  - BICONVEX_TOL
```
public static final String BICONVEX_TOL
```
    The threshold of the ratio between the (sum(new_w - old_w) + sum(new_u - old_u)) / (sum(old_u) + sum(old_w)) i.e. some notion of normalised changed of the paramters. Defaults to 0.01
    
    See Also:
    
    Constant Field Values
  - LAMBDA_W
```
public static final String LAMBDA_W
```
    The parameter of the regulariser for W, defaults to LAMBDA
    
    See Also:
    
    Constant Field Values
  - LAMBDA_U
```
public static final String LAMBDA_U
```
    The parameter of the regulariser for U, defaults to LAMBDA
    
    See Also:
    
    Constant Field Values
  - LAMBDA
```
public static final String LAMBDA
```
    The parameter of the regulariser for both W and U
    
    See Also:
    
    Constant Field Values
  - ETA0_U
```
public static final String ETA0_U
```
    The weighting of the subgradient of U, weighted down each ETASTEPS number of iterations of the biconvex scheme, defaults to 0.05
    
    See Also:
    
    Constant Field Values
  - ETA0_W
```
public static final String ETA0_W
```
    The weighting of the subgradient of W, weighted down each ETASTEPS number of iterations of the biconvex scheme, defaults to 0.05
    
    See Also:
    
    Constant Field Values
  - ETA0_BIAS
```
public static final String ETA0_BIAS
```
    The weighting of the subgradient of BIAS, weighted down each ETASTEPS number of iterations of the biconvex scheme, defaults to eta0 (0.05)
    
    See Also:
    
    Constant Field Values
  - LOSS
```
public static final String LOSS
```
    The loss function, defaults to SquareMissingLossFunction
    
    See Also:
    
    Constant Field Values
  - REGUL
```
public static final String REGUL
```
    The regularisation function, defaults to L1L2Regulariser
    
    See Also:
    
    Constant Field Values
  - ETASTEPS
```
public static final String ETASTEPS
```
    The steps at which point the eta parameter is reduced, defaults to 3
    
    See Also:
    
    Constant Field Values
  - FORCE_SPARCITY
```
public static final String FORCE_SPARCITY
```
    Should all parameter matricies be held SparseMatrix instances and therefore remain sparse. Forces a copy but could save a lot.
    
    See Also:
    
    Constant Field Values
  - DAMPENING
```
public static final String DAMPENING
```
    The value of w, u and beta are updated each time data is added s.t. w = w * (1.0 - DAM PENING). The default value is 0
    
    See Also:
    
    Constant Field Values
  - Z_STANDARDISE
```
public static final String Z_STANDARDISE
```
    Whether the Vprime and Dprime matrices should be zscore standardised
    
    See Also:
    
    Constant Field Values
  - ETA_GAMMA
```
public static final String ETA_GAMMA
```
    The ammount by which ETA is made to increase each iteration
    
    See Also:
    
    Constant Field Values
- Constructor Detail
  - BilinearLearnerParameters
```
public BilinearLearnerParameters()
```
    sets up the defaults
- Method Detail
  - writeASCII
```
public void writeASCII(PrintWriter out)
                throws IOException
```
    Description copied from interface: WriteableASCII
    
    Write the content of this as ascii to out.
    
    Specified by:
    
    writeASCII in interface WriteableASCII
    
    Parameters:
    
    out - sink to write to
    
    Throws:
    
    IOException - an error writing to out
  - toString
```
public String toString()
```
    Overrides:
    
    toString in class AbstractMap<String,Object>
  - asciiHeader
```
public String asciiHeader()
```
    Description copied from interface: WriteableASCII
    
    Header for ascii output. Will be automatically written by IOUtils when using writeASCII().
    
    Specified by:
    
    asciiHeader in interface WriteableASCII
    
    Returns:
    
    header
  - clone
```
public BilinearLearnerParameters clone()
```
    Overrides:
    
    clone in class HashMap<String,Object>

Class BilinearLearnerParameters

Nested Class Summary

Nested classes/interfaces inherited from class org.openimaj.ml.linear.learner.LearningParameters

Nested classes/interfaces inherited from class java.util.AbstractMap

Nested classes/interfaces inherited from interface java.util.Map

Field Summary

Constructor Summary

Method Summary

Methods inherited from class org.openimaj.ml.linear.learner.LearningParameters

Methods inherited from class java.util.HashMap

Methods inherited from class java.util.AbstractMap

Methods inherited from class java.lang.Object

Methods inherited from interface java.util.Map

Field Detail

BIAS

SEED

WINITSTRAT

UINITSTRAT

EXPANDEDWINITSTRAT

EXPANDEDUINITSTRAT

BIASINITSTRAT

BICONVEX_MAXITER

BICONVEX_TOL

LAMBDA_W

LAMBDA_U

LAMBDA

ETA0_U

ETA0_W

ETA0_BIAS

LOSS

REGUL

ETASTEPS

FORCE_SPARCITY

DAMPENING

Z_STANDARDISE

ETA_GAMMA

Constructor Detail

BilinearLearnerParameters

Method Detail

writeASCII

toString

asciiHeader

clone