Implemented sd_ratio as quality metric
#2146
Conversation
for more information, see https://pre-commit.ci
for more information, see https://pre-commit.ci
| @@ -1365,3 +1365,57 @@ def _compute_rp_violations_numba(nb_rp_violations, spike_trains, spike_clusters, | |||
| spike_train = spike_trains[spike_clusters == i] | |||
| n_v = _compute_nb_violations_numba(spike_train, t_r) | |||
| nb_rp_violations[i] += n_v | |||
|
|
|||
|
|
|||
| def compute_SD_test( | |||
There was a problem hiding this comment.
| def compute_SD_test( | |
| def compute_sd_ratio( |
I'm not super happy with the name. What abut sd_ratio?
There was a problem hiding this comment.
Yeah I'm fine with that name!
sd_ratio or std_ratio?
DradeAW
changed the title
sd_ratio as quality metric
|
@zm711 Thanks for the feedback! I'll make the changes after my lunch! |
for more information, see https://pre-commit.ci
for more information, see https://pre-commit.ci
|
Ok I like the state of the PR as it is now. A few things that are not perfect:
|
|
Could you also do a big favor with this PR and fix the order of |
| warnings.warn( | ||
| "The `sd_ratio` metric cannot work with a recordless WaveformExtractor object" | ||
| "SD ratio metric will be set to NaN" | ||
| ) |
There was a problem hiding this comment.
we have to decide a global behavior for this. I would say lets raise no ?
| spike_train = wvf_extractor.sorting.get_unit_spike_train(unit_id, segment_index=segment_index).astype( | ||
| np.int64 | ||
| ) | ||
| censored_indices = _find_duplicated_spikes_keep_first_iterative(spike_train, censored_period) |
There was a problem hiding this comment.
Not sure this is a good idea this cross dependency in this way between curation and qualitymetrics.
There was a problem hiding this comment.
What would you suggest as an alternative?
There was a problem hiding this comment.
that's ok for now, but we can discuss higher-level hierarchy later
|
Hi Aurelien. |
|
Yes this is different than the 2002 paper (in multiple ways), but I still cited it because it was the first one to write down the idea. We can talk about it. I found this implementation easier, but you are right there may be more information available by using the whole waveform (although it makes drift correction and censored period trickier). |
|
The test is failing because of a bug that is solved by #2220 |
A "new" metric I've been working on.
I hope to implement it fast enough for it to be part of v0.99