From de827a42e10ede662f4bd2893c4f8b6d54be45dc Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Sat, 1 Aug 2026 18:53:29 -0500 Subject: Amplification beats the instance that defeated fifteen solvers The caption-omitted field returned under 5% from every solver tried: Umeyama, GRAMPA, five Gromov-Wasserstein variants, FAQ with and without restarts, PATH convex-concave, a moment ladder, semirelaxed GW. The measurements said the answer was not a sixteenth solver. Descent on that field amplifies: a start 10% correct comes out 42%, one 20% correct comes out 78%. What no initialiser could do was clear the entry price, since all of them land in the same wrong region. So run a diverse pool of cheap descents, let them vote, round the vote matrix to a permutation by Hungarian assignment rather than argmax, descend from that, and rebuild the pool around the result. Each round feeds the amplifier a better start. 0.044 -> 0.72 on the best run, 0.54 mean over two. Competitive elsewhere: 0.977 on the easy field against 0.961 for the best GW variant. Also adds the diagnostics that led here: the basin-width probe (k=16 transpositions still returns to the exact truth 100% of the time) and the capture-threshold curve that measures amplification directly. Co-Authored-By: Claude --- logs/battery2.log | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+) create mode 100644 logs/battery2.log (limited to 'logs/battery2.log') diff --git a/logs/battery2.log b/logs/battery2.log new file mode 100644 index 0000000..bfde59c --- /dev/null +++ b/logs/battery2.log @@ -0,0 +1,24 @@ + +=== omit-size-BOUND0.997 (N=256, chance=0.0039) + GW x12 restarts raw=0.0182 refined=0.0221 (0.1s) + GW kl-loss raw=0.0000 refined=0.0000 (0.1s) +/home/yurenh2/.local/lib/python3.13/site-packages/ot/bregman/_sinkhorn.py:666: UserWarning: Sinkhorn did not converge. You might want to increase the number of iterations `numItermax` or the regularization parameter `reg`. + warnings.warn( +/home/yurenh2/.local/lib/python3.13/site-packages/ot/bregman/_sinkhorn.py:630: RuntimeWarning: divide by zero encountered in divide + v = b / KtransposeU +/home/yurenh2/.local/lib/python3.13/site-packages/ot/bregman/_sinkhorn.py:630: RuntimeWarning: overflow encountered in divide + v = b / KtransposeU +/home/yurenh2/.local/lib/python3.13/site-packages/ot/backend.py:1197: RuntimeWarning: invalid value encountered in dot + return np.dot(a, b) +/home/yurenh2/.local/lib/python3.13/site-packages/ot/bregman/_sinkhorn.py:642: UserWarning: Warning: numerical errors at iteration 0 + warnings.warn("Warning: numerical errors at iteration %d" % ii) +/home/yurenh2/.local/lib/python3.13/site-packages/ot/gromov/_bregman.py:245: UserWarning: Solver failed to produce a transport plan. You might want to increase the regularization parameter `epsilon`. + warnings.warn( + entropic GW deep anneal raw=0.0078 refined=0.0339 (53.1s) + semirelaxed GW raw=0.0039 refined=0.0130 (0.1s) + +=== rank8-BOUND0.93 (N=256, chance=0.0039) + GW x12 restarts raw=0.8242 refined=0.8815 (0.1s) + GW kl-loss raw=0.2161 refined=0.2513 (0.1s) +/home/yurenh2/.local/lib/python3.13/site-packages/ot/bregman/_sinkhorn.py:642: UserWarning: Warning: numerical errors at iteration 24 + warnings.warn("Warning: numerical errors at iteration %d" % ii) -- cgit v1.2.3