Description/summary not set
Other contact details:
Not specified
Interests:
Feature Selection, Microarray data
Field/Industry:
Not specified
Occupation/Role(s):
Not specified
Organisation(s):
Not specified
Note: some items may not be visible to you, due to viewing permissions.
Contents (click to expand/contract)
Is Parameter Optimization Useful
(1)
This process compares two 10x10fold cross-validations with linear SVM for classification. One use the heuristic parameter choice C=1/avg(K(x,x)) and the other uses an internal 5fold cross-validation to tune the parameter C. Afterwards both performance values are compared via ANOVA and T-test to check whether there is a significant difference between both performance values.
The behaviour of the optimized 10x10 XV is similar to that of the experiment 10x10fold_XV_with_optimizedSVM_classificat...
Created: 2010-10-28
| Last updated: 2010-10-28
10x10fold_XV_with_optimizedSVM_classification
(1)
This process performs 10 round of 10fold cross-validation for classification with a linear SVM. In each traning-phase the parameters C of the linear SVM is tuned by a 5fold cross-validation.
The repetions are useful for estimating the real performance on a small sample high variance data set, e.g. microarray data.
Created: 2010-10-28
Find best feature selection method.
(1)
This processes searches for the best method for feature selection and the best feature subset size. Inside a parameter loop the number of selected features and the feature selection methods are evaluated with a 10fold cross-validation. The Feature Selection Extension ( https://sourceforge.net/projects/rm-featselext ) is required.
BUGFIX: Please note that the plugin downloaded before 5th November contains a bug. Due to a bug in RM the TopK-operator sorted the features in the wrong direction. ...
Created: 2010-10-19
| Last updated: 2010-12-10
Recursive Feature Elimination with fully o...
(1)
The Process shows how you can use the RecursiveFeatureElimination-Operator of the Feature Selection Extension ( https://sourceforge.net/projects/rm-featselext) to select relevant features from a highdimensional dataset.
Created: 2010-10-19
| Last updated: 2010-12-10