Closed jdrusso closed 1 year ago
Merging #41 (1547aa4) into develop (69fd248) will decrease coverage by
0.31%
. The diff coverage is24.48%
.:exclamation: Current head 1547aa4 differs from pull request most recent head 1b7e5b8. Consider uploading reports for the commit 1b7e5b8 to get more accurate results
@@ Coverage Diff @@
## develop #41 +/- ##
===========================================
- Coverage 53.60% 53.29% -0.31%
===========================================
Files 25 25
Lines 3871 3916 +45
===========================================
+ Hits 2075 2087 +12
- Misses 1796 1829 +33
Impacted Files | Coverage Δ | |
---|---|---|
msm_we/_hamsm/_plotting.py | 5.26% <0.00%> (-0.56%) |
:arrow_down: |
msm_we/_hamsm/_analysis.py | 35.61% <9.52%> (-1.56%) |
:arrow_down: |
msm_we/_hamsm/_clustering.py | 54.15% <100.00%> (+0.77%) |
:arrow_up: |
:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more
Sometimes, it appears that repeated restarting from poor models can drive the system to a higher-flux steady state
One heuristic for detecting this is "overccorrection" in the flux profiles -- i.e., it should start out higher near the source, and lower near the target, and flatten out. But, when these "bad restarts" happen, it seems to consistently drive the flux profile past flat, to higher near the target, and lower near the source.
We can detect this by doing a curve fit, and looking for when the slope is "wrong" in that way