In random select of weak models, models may be included that return NaN (if they're sufficiently bad) which ruins the stacked model (it cannot handel NaNs).
Solution: filter out worst of the weak models (if they return NaN), or make stacking regressor handle poor consituent models
Related to Issue #10
In random select of weak models, models may be included that return NaN (if they're sufficiently bad) which ruins the stacked model (it cannot handel NaNs).
Solution: filter out worst of the weak models (if they return NaN), or make stacking regressor handle poor consituent models