-
Notifications
You must be signed in to change notification settings - Fork 1
feature selection
Austin Richardson edited this page Jan 22, 2015
·
1 revision
I have decided to implement three types of feature selection:
- Analysis of Variance (single feature).
- Decision Tree + ANOVA (1+ features).
- Decision Tree + ANOVA culled by feature location (1+ features, considers homology).
Class labels are determined by user but can be automatically assigned to either (a) taxonomic nomenclature or (b) label-free clustering.
All will be coupled with cross-validation in order to estimate variance and generate feature importance plots.
I need to implement these as IPython notebooks first in order to benchmark them.