when features go missing the system should use the "next best" points, which we can find using the sorted list of features at the initialization.
We also need to evaluate when we stop using those features (for example if the weight function falls below a certain threshold).
when features go missing the system should use the "next best" points, which we can find using the sorted list of features at the initialization. We also need to evaluate when we stop using those features (for example if the weight function falls below a certain threshold).