Proposals:Refactoring Statistics Framework 2007 Background: Difference between revisions

From KitwarePublic
Jump to navigationJump to search
No edit summary
No edit summary
 
(4 intermediate revisions by the same user not shown)
Line 7: Line 7:
## Image
## Image
## Data points
## Data points
#Membership function
#Membership models
## Distance functions
## Can be manually set or automatically generated from the sample data
## Can be manually set or automatically generated from the sample data
## Estimators are available to generate membership functions ( ImageModelEstimatorBase, ImageGuassianModelEstimator,ExpectationMaximizationMixtureModelEstimator  )
## Estimators are available to generate membership functions ( ImageModelEstimatorBase, ImageGuassianModelEstimator,ExpectationMaximizationMixtureModelEstimator  )
## Some classes are named with Estimator suffix but they do more than just estimating membership functions
## Some classes are named with Estimator suffix but they do more than just estimating membership functions
###  itkKdTreeBasedKmeansEstimator
###  itkKdTreeBasedKmeansEstimator
# Decison Rule (Classifier ): such as
# Distance functions
# Decision Rules
# Classifiers


 
Note:
Typical scenario
# Classifiers provide interface that integrates all the other components. Classifiers provide a common framework
# Use an estimator to generate class models for input data.
# ITK also contains classes which combine specific types of the different components into one Huge framework such as itkScalarImageKmeansImageFilter and itkBayesianClassifierImageFilter.
# Use the generated class models, distance function and a decision rule to determine which class your
## itkScalarImageKmeansImageFilter: EuclideanDistance, KdTreeBasedKmeansEstimator, SampleClassifier, MinimumDecisionRule
input belongs to.
## itkBayesianClassifierImageFilter: Bayesian Estimator, MaximumDecisionRule

Latest revision as of 21:19, 16 July 2008

DudaClassifier.png StatisticalClassificationFramework.png


The main components of a classification framework are

  1. Input
    1. Image
    2. Data points
  2. Membership models
    1. Can be manually set or automatically generated from the sample data
    2. Estimators are available to generate membership functions ( ImageModelEstimatorBase, ImageGuassianModelEstimator,ExpectationMaximizationMixtureModelEstimator )
    3. Some classes are named with Estimator suffix but they do more than just estimating membership functions
      1. itkKdTreeBasedKmeansEstimator
  3. Distance functions
  4. Decision Rules
  5. Classifiers

Note:

  1. Classifiers provide interface that integrates all the other components. Classifiers provide a common framework
  2. ITK also contains classes which combine specific types of the different components into one Huge framework such as itkScalarImageKmeansImageFilter and itkBayesianClassifierImageFilter.
    1. itkScalarImageKmeansImageFilter: EuclideanDistance, KdTreeBasedKmeansEstimator, SampleClassifier, MinimumDecisionRule
    2. itkBayesianClassifierImageFilter: Bayesian Estimator, MaximumDecisionRule
Retrieved from "https://public.kitware.com/Wiki/index.php?title=Proposals:Refactoring_Statistics_Framework_2007_Background&oldid=12663"