It may be useful to allow a single "global" threshold (or maybe a fraction of sub-theshold features) across amp_fract, amp_consist, period_consist, monotonicity - assuming there is an appropriate way normalize / weight these features together.
For example, a cycle that is sub-threshold for all 4/4 features is much more likely to be non-burst than a cycle that is sub-threshold for 1/4 features. But with the current module, both cycles would be considered non-burst. This leads to headaches when tuning burst detection - espically with noisy sims.
It may be useful to allow a single "global" threshold (or maybe a fraction of sub-theshold features) across amp_fract, amp_consist, period_consist, monotonicity - assuming there is an appropriate way normalize / weight these features together.
For example, a cycle that is sub-threshold for all 4/4 features is much more likely to be non-burst than a cycle that is sub-threshold for 1/4 features. But with the current module, both cycles would be considered non-burst. This leads to headaches when tuning burst detection - espically with noisy sims.