Explicit Control of Feature Relevance and Selection Stability Through Pareto Optimality

I Increasing stability . Ensemble feature selection : selects features that are selected the most accross different selection runs. . Instance weighting : weights training instances according to their importance to feature evaluation. . Model selection: takes stability into account in the fitting of the meta-parameters. ⇒ No fine control of the accuracy-stability trade-off. I Stability measure [1]