diff options
| -rw-r--r-- | figs/most_squares_complex.pdf | bin | 0 -> 12160 bytes | |||
| -rw-r--r-- | figs/most_squares_complexity.pdf | bin | 51036 -> 51036 bytes | |||
| -rw-r--r-- | figs/most_squares_stability.pdf | bin | 41384 -> 41384 bytes | |||
| -rw-r--r-- | marginal.tex | 2 |
4 files changed, 1 insertions, 1 deletions
diff --git a/figs/most_squares_complex.pdf b/figs/most_squares_complex.pdf Binary files differnew file mode 100644 index 0000000..dbfc896 --- /dev/null +++ b/figs/most_squares_complex.pdf diff --git a/figs/most_squares_complexity.pdf b/figs/most_squares_complexity.pdf Binary files differindex 6f54dc4..7e07630 100644 --- a/figs/most_squares_complexity.pdf +++ b/figs/most_squares_complexity.pdf diff --git a/figs/most_squares_stability.pdf b/figs/most_squares_stability.pdf Binary files differindex 9f15559..a96b62e 100644 --- a/figs/most_squares_stability.pdf +++ b/figs/most_squares_stability.pdf diff --git a/marginal.tex b/marginal.tex index dd7bf68..b95872b 100644 --- a/marginal.tex +++ b/marginal.tex @@ -109,7 +109,7 @@ further looking for a sign that the zero eigenvalue lies at the edge of a continuous spectrum, we enforce the condition that the spectrum has a pseudogap, and is therefore marginal. We demonstrate the method on the spherical spin glasses, where it is unnecessary but instructive, and on extensions of -the spherical models with non-GOE Hessians where the technique is more useful. +the spherical models where the technique is more useful. In a companion paper, we compare the marginal complexity with the performance of gradient descent and approximate message passing algorithms \cite{Kent-Dobias_2024_Algorithm-independent}. |