| MATLAB File Help: dml.filterer | View code for dml.filterer | Default Topics |
filterer filtering approach to feature selection.
DESCRIPTION
The filterer produces an ordering of the features based on a univariate measure
and uses that ordering to greedily determine the 'optimal' feature set
or to select the best m features.
If filter is a string then it calls rankfeatures in the bioinformatics
toolbox. If it is a function it will just evaluate that function on
each feature and the output.
EXAMPLE
dml.filterer('validator',dml.crossvalidator('mva',dml.naive,'stat','accuracy'))
DEVELOPER
Marcel van Gerven (m.vangerven@donders.ru.nl)| Superclasses | dml.method |
| Sealed | false |
| Construct on load | false |
| filterer | filtering approach to feature selection. |
| criterion | the value of the validator for feature 1:N |
| filter | the used filter |
| indims | dimensions of the input data (excluding the trial dim and time dim in time series data) |
| maxfeatures | maximum number of used features |
| order | the ordering of the features according to value |
| restart | when false, starts at the previously learned parameters; needed for online learning and grid search |
| subset | subset of selected features |
| validator | the used crossvalidator |
| value | the value of the filter for each feature |
| verbose | whether or not to generate diagnostic output |
| model | return logical array with ones indicating selected features | |
| test | ||
| train | multiple datasets |