Closed aalexmmaldonado closed 6 months ago
So, there are indeed negative weights when we have (i, i) edges. They are essentially zero, but due to NumPy floats, they are on the order of $-1 \times 10^{-16}$. We will set all diagonals of correlation_matrix
to 0.0
in run_wisp()
.
All bugs for serial processing have been fixed at this point. I left the new test running for 30 minutes, and it would not finish. We must debug and profile the code to see the most significant contributions.
We need to merge so we can debug effectively. Will make another PR. #1 is still not fixed.
These tests work on my local computer. Something is going on with the workflow.