SparseICA

All Classes

SEARCH:

Summary:
Nested |
Field |
Constr |
Method

Detail:
Field |
Constr |
Method

java.lang.Object
- samer.tools.AnonymousTask
- - samer.tools.NamedTask
  - - samer.models.SparseICA

All Implemented Interfaces:: Model, Task

public class SparseICA
extends NamedTask
implements Model

See Also:: Serialized Form

Nested Class Summary

Nested Classes
Modifier and Type	Class and Description
`class`	`SparseICA.ON2Trainer` This trainer uses an O(N^2) run step and an O(N^2) flush.
`class`	`SparseICA.ON3Trainer` This trainer saves on an O(N^2) step during accumulation, at the expense of an O(N^3) flush.

Nested classes/interfaces inherited from interface samer.models.Model
Model.Trainer

Field Summary
- Fields inherited from class samer.tools.NamedTask
  node

Constructor Summary

Constructors
Constructor and Description

SparseICA(int N)

SparseICA(Vec input)

Constructors
Constructor and Description
`SparseICA(int N)`
`SparseICA(Vec input)`

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`void`	`compute()` contract is that getEnergy and getGradient must return correct values for current x after infer and compute has been called, but not necessarily before.
`void`	`dispose()`
`void`	`execute(java.lang.String cmd, Environment env)`
`Functionx`	`functionx()` This presents a more functional interface to the model so that it can be driven by an optimiser.
`Model.Trainer`	`getAltTrainer()`
`void`	`getCommands(Registry r)`
`double`	`getEnergy()` return E = -log p(x)
`double[]`	`getGradient()` return dE/dx
`Model`	`getOutputModel()`
`int`	`getSize()` return size of vector this model expects
`Model.Trainer`	`getTrainer()`
`SparseMatrix`	`getWeightMatrix()`
`void`	`infer()` should infer values latent variables
`void`	`init1(Matrix R, int M)` x is an N-dim vector.
`void`	`init2(Matrix R)` This version only builds disjoint butterfly pairs, ie only 2N edges per layer Need a version that builds bigger local modules.
`VVector`	`output()`
`void`	`run()`
`void`	`setInput(Vec input)` model should begin observing this vector
`void`	`setOutputModel(Model m)`
`void`	`starting()`
`void`	`stopping()`

Methods inherited from class samer.tools.NamedTask
getNode, toString

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

- Constructor Detail
  - SparseICA
```
public SparseICA(int N)
```
  - SparseICA
```
public SparseICA(Vec input)
```
- Method Detail
  - getSize
```
public int getSize()
```
    Description copied from interface: Model
    
    return size of vector this model expects
    
    Specified by:
    
    getSize in interface Model
  - output
```
public VVector output()
```
  - getOutputModel
```
public Model getOutputModel()
```
  - setOutputModel
```
public void setOutputModel(Model m)
```
  - setInput
```
public void setInput(Vec input)
```
    Description copied from interface: Model
    
    model should begin observing this vector
    
    Specified by:
    
    setInput in interface Model
  - getWeightMatrix
```
public SparseMatrix getWeightMatrix()
```
  - dispose
```
public void dispose()
```
    Specified by:
    
    dispose in interface Model
    
    Specified by:
    
    dispose in interface Task
    
    Overrides:
    
    dispose in class NamedTask
  - infer
```
public void infer()
```
    Description copied from interface: Model
    
    should infer values latent variables
    
    Specified by:
    
    infer in interface Model
  - compute
```
public void compute()
```
    Description copied from interface: Model
    
    contract is that getEnergy and getGradient must return correct values for current x after infer and compute has been called, but not necessarily before. This is to give model an opportunity to cache values of energy and gradient to avoid repeated computations.
    
    Specified by:
    
    compute in interface Model
  - getEnergy
```
public double getEnergy()
```
    Description copied from interface: Model
    
    return E = -log p(x)
    
    Specified by:
    
    getEnergy in interface Model
  - getGradient
```
public double[] getGradient()
```
    Description copied from interface: Model
    
    return dE/dx
    
    Specified by:
    
    getGradient in interface Model
  - functionx
```
public Functionx functionx()
```
    Description copied from interface: Model
    
    This presents a more functional interface to the model so that it can be driven by an optimiser. See classes Functionx and MinimiserBase in package samer.maths.opt.
    
    Specified by:
    
    functionx in interface Model
  - starting
```
public void starting()
```
    Specified by:
    
    starting in interface Task
    
    Overrides:
    
    starting in class AnonymousTask
  - stopping
```
public void stopping()
```
    Specified by:
    
    stopping in interface Task
    
    Overrides:
    
    stopping in class AnonymousTask
  - run
```
public void run()
```
    Specified by:
    
    run in interface Task
  - getCommands
```
public void getCommands(Registry r)
```
  - execute
```
public void execute(java.lang.String cmd,
                    Environment env)
```
  - getTrainer
```
public Model.Trainer getTrainer()
```
  - getAltTrainer
```
public Model.Trainer getAltTrainer()
```
  - init1
```
public void init1(Matrix R,
                  int M)
```
    x is an N-dim vector. R is an N by N symmetric similarity matrix. M is number of unit pairs (butterflies) build into ICA model
  - init2
```
public void init2(Matrix R)
```
    This version only builds disjoint butterfly pairs, ie only 2N edges per layer Need a version that builds bigger local modules.

Skip navigation links

All Classes

Summary:
Nested |
Field |
Constr |
Method

Detail:
Field |
Constr |
Method

Java-Scheme Laboratory

Class SparseICA

Nested Class Summary

Nested classes/interfaces inherited from interface samer.models.Model

Field Summary

Fields inherited from class samer.tools.NamedTask

Constructor Summary

Method Summary

Methods inherited from class samer.tools.NamedTask

Methods inherited from class java.lang.Object

Constructor Detail

SparseICA

Method Detail

getSize

output

getOutputModel

setOutputModel

setInput

getWeightMatrix

dispose

infer

compute

getEnergy

getGradient

functionx

starting

stopping

run

getCommands

execute

getTrainer

getAltTrainer

init1

init2