Skip to main content
24 events
when toggle format what by license comment
Dec 4, 2023 at 11:30 history edited YCor CC BY-SA 4.0
formatting
Dec 4, 2023 at 11:13 history edited loup blanc CC BY-SA 4.0
added 597 characters in body
Oct 20, 2023 at 21:56 answer added Dan Piponi timeline score: 5
Oct 20, 2023 at 16:38 comment added Michael Hardy The probability distribution of this determinant seems to have tails so fat that the sample mean probabily has a much larger variance than the sample median.
Oct 20, 2023 at 15:39 answer added Mykola Pochekai timeline score: 19
Oct 20, 2023 at 9:58 comment added loup blanc @Abdelmalek Abdesselam ; thanks ; great proof.
Oct 18, 2023 at 21:58 comment added Dan Piponi Going back to my example of the sum of $m$ squares of $n\times n$ matrices. By an argument related to the classical matchings-via-determinants one, each non-zero term contributing to the determinant corresponds to a permutation where each cycle is given 1 of m colors and it contributes precisely 1. Eg. for n=3, m=2 we have 1 cycle conjugate to (1)(2)(3) contributing 2^3, 3 cycles conjugate to (1)(23) contributing 2^2 and 2 cycles conjugate to (123) contributing 2^1=8+3*4+2*2=24=(3+1)!
Oct 18, 2023 at 20:29 answer added Abdelmalek Abdesselam timeline score: 21
Oct 18, 2023 at 20:14 comment added Denis Serre Another instance of non-commutative probabilities ... :)
Oct 18, 2023 at 20:02 comment added Carlo Beenakker @DanPiponi -- FWIW, my attempt at an answer, which I was not able to complete for arbitrary $n$, does directly prove your equation for $n=2$ and arbitrary $m$ [when the determinant equals $m(m+1)$]
Oct 18, 2023 at 17:32 comment added Dan Piponi More generally I guess $E[\det(X_1^2+X_2^2+\ldots+X_m^2)]=(m+n-1)!/(m-1)!$ for independent matrices $X_i$ of the same type.
Oct 18, 2023 at 16:00 comment added loup blanc @YCor, yes you are right.
Oct 18, 2023 at 15:57 history edited loup blanc CC BY-SA 4.0
added 2 characters in body
Oct 18, 2023 at 13:22 comment added YCor The tag gaussian already exists, no need to create normal-law. Unless you mean something else that Gaussian/Normal distribution (but what?).
Oct 18, 2023 at 13:21 history edited YCor
edited tags
Oct 18, 2023 at 13:06 answer added Carlo Beenakker timeline score: 7
Oct 18, 2023 at 11:45 comment added Timothy Chow Wayback Machine link to Moo K. Chung's paper: On Expected Gaussian Random Determinants.
Oct 18, 2023 at 8:19 comment added Carlo Beenakker @TimothyBudd --- yes, apologies, $\mathbb{E}[\det(X+X^\top)]=(-1)^m (2m)!/m!$ (this is stated correctly by Chung, it's my mistake)
Oct 18, 2023 at 7:24 comment added Timothy Budd @CarloBeenakker I suppose in your formula $Y$ is supposed to be the transpose of $X$? If $X$ and $Y$ are independent, the determinant vanishes.
Oct 17, 2023 at 22:58 history edited YCor CC BY-SA 4.0
formatting
Oct 17, 2023 at 21:22 comment added Joseph Van Name I formulated and verified a conjecture similar to Conjecture 1 for quaternionic matrices. If $A$ is an $m\times n$-quaternionic matrix, then let $\chi_A$ be the corresponding $2m\times 2n$-complex matrix. Then whenever $A$ is a random $n\times n$-quaternionic matrices, we experimentally have $E(\det(\chi_A))=n!$. Here, the entries $v$ in $A$ are Gaussian with independent with mean $0$ and where if $\alpha$ is a unit quaternion, then $\text{Var}(\text{Re}(\alpha v))=1/4$.
Oct 17, 2023 at 20:50 history edited Carlo Beenakker CC BY-SA 4.0
somewhat more informative title
Oct 17, 2023 at 20:48 comment added Carlo Beenakker interesting conjecture; a similar identity, $\mathbb{E}[\det(X+Y)]=(-1)^m (2m)!/m!$ for $n=2m$, was proven by M.K. Chung
Oct 17, 2023 at 19:31 history asked loup blanc CC BY-SA 4.0