Closed damb closed 2 years ago
@mmesim, I added some more MLx
related tests.
As discussed, could you please compute the results from the integration-tests
magnitude/MLx/single-detector-single-stream-0003
magnitude/MLx/single-detector-multi-stream-0003
magnitude/MLx/single-detector-multi-stream-0004
independently? (You'll find the test data in the src/apps/cc/test/data/integration/
directory.) Note that I'm going to look into the latter two tests, again by myself, as well. The MLx
magnitudes seem somehow inconsistent.
Thanks.
OT: This is detection related (and not MLx
amplitude/magnitude related). In the context of the integration-tests
magnitude/MLx/single-detector-multi-stream-0004
I noted an interesting behavior when increasing the "arrivalOffsetThreshold"
to e.g. 0.2
, i.e. instead of detecting the template origin (2019-11-05T05:23:47.640489Z
), only, an additional detection is emitted (2019-11-05T05:23:29.035489Z
).
Attached the notebook used to validate the tests. Credits to @mmesim.
Note that the actual implementation uses a slightly different order when it comes to preprocessing the data. E.g. the data is filtered before it is deconvoluted.
Closes #75.