Not signed in (Sign In)

Start a new discussion

Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

  • Sign in using OpenID

Site Tag Cloud

2-category 2-category-theory abelian-categories adjoint algebra algebraic algebraic-geometry algebraic-topology analysis analytic-geometry arithmetic arithmetic-geometry bundles calculus categorical categories category category-theory chern-weil-theory cohesion cohesive-homotopy-type-theory cohomology colimits combinatorics complex-geometry computable-mathematics computer-science constructive cosmology deformation-theory descent diagrams differential differential-cohomology differential-equations differential-geometry digraphs duality elliptic-cohomology enriched fibration finite foundations functional-analysis functor galois-theory gauge-theory gebra geometric-quantization geometry graph graphs gravity grothendieck group group-theory harmonic-analysis higher higher-algebra higher-category-theory higher-differential-geometry higher-geometry higher-lie-theory higher-topos-theory homological homological-algebra homotopy homotopy-theory homotopy-type-theory index-theory integration integration-theory internal-categories k-theory lie lie-theory limit limits linear linear-algebra locale localization logic manifolds mathematics measure-theory modal modal-logic model model-category-theory monads monoidal monoidal-category-theory morphism motives motivic-cohomology nforum nlab noncommutative noncommutative-geometry number number-theory of operads operator operator-algebra order-theory pages pasting philosophy physics pro-object probability probability-theory quantization quantum quantum-field quantum-field-theory quantum-mechanics quantum-physics quantum-theory question representation representation-theory riemannian-geometry scheme schemes set set-theory sheaf simplicial space spin-geometry stable-homotopy-theory string string-theory superalgebra supergeometry svg symplectic-geometry synthetic-differential-geometry terminology theory topology topos topos-theory type type-theory universal variational-calculus

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

Welcome to nForum
If you want to take part in these discussions either sign in now (if you have an account), apply for one now (if you don't).
    • CommentRowNumber1.
    • CommentAuthorUrs
    • CommentTimeApr 29th 2021

    Page created, but author did not leave any comments.

    v1, current

    • CommentRowNumber2.
    • CommentAuthorUrs
    • CommentTimeMay 9th 2021

    added under “Properties” (here) brief statement of and pointer to the hook-content formula.

    diff, v4, current

    • CommentRowNumber3.
    • CommentAuthorUrs
    • CommentTimeMay 15th 2021
    • (edited May 15th 2021)

    I have expanded out the statement of the definition, making everything fully explicit; also defining standard Young diagrams.

    Also added a minimum of an Idea-section.

    diff, v5, current

    • CommentRowNumber4.
    • CommentAuthorUrs
    • CommentTimeMay 15th 2021
    • (edited May 15th 2021)

    In the discussion of the relation to Schur polynomials I have made more explicit (here) what the formula reduces to for Schur polynomials of a finite number of variables

    diff, v5, current

    • CommentRowNumber5.
    • CommentAuthorUrs
    • CommentTimeMay 19th 2021

    added this pointer:

    diff, v6, current

    • CommentRowNumber6.
    • CommentAuthorDavid_Corfield
    • CommentTimeMay 29th 2021

    Removed a 1/21/2 which shouldn’t be there.

    diff, v12, current

    • CommentRowNumber7.
    • CommentAuthorUrs
    • CommentTimeJun 1st 2021
    • (edited Jun 1st 2021)

    Now I have put it all together and summed up (here) the full asymptotic expansion of the number of height-bounded standard Young tableaux. Currently I get this expression:

    ln(|sYT n(N)|) nNnln(N)14N 2ln(n)+14Nln(n) nN+12N 2ln(N)(38+ln(2)4)N 214Nln(N)+ln(2)2N16ln(N) nN+(2ζ (1)ln(2)6lnG(1/2)) \begin{aligned} & ln \big( \left\vert sYT_n(N) \right\vert \big) \\ & \; \underset{ { n \to \infty } \atop { N \to \infty } }{\sim}\; n ln(N) - \tfrac{1}{4}N^2 ln(n) + \tfrac{1}{4}N ln(n) \\ & \phantom{ \underset{ { n \to \infty } \atop { N \to \infty } }{\sim}\; } + \tfrac{1}{2} N^2 ln(N) - \left( \tfrac{3}{8} + \tfrac{ln(2)}{4} \right) N^2 - \tfrac{1}{4} N ln(N) + \tfrac{ln(2)}{2} N - \tfrac{1}{6} ln(N) \\ & \; \phantom{ \underset{ { n \to \infty } \atop { N \to \infty } }{\sim}\; } + \big( 2 \zeta^'(-1) - \tfrac{ln(2)}{6} - ln G(1/2) \big) \end{aligned}

    I have checked factors a few times, but it won’t hurt to check again… (and thanks for catching a wrong factor in #6!)

    diff, v16, current

    • CommentRowNumber8.
    • CommentAuthorDavid_Corfield
    • CommentTimeJun 1st 2021

    Are there issues about taking limits in an order?

    Next would be to get an approximation for min-entropy.

    • CommentRowNumber9.
    • CommentAuthorUrs
    • CommentTimeJun 1st 2021
    • (edited Jun 1st 2021)

    Are there issues about taking limits in an order?

    I was wondering about that, too. Maybe this needs more thinking.

    Incidentally, I was just turning again to Mkrtchyan 14. Its main Theorem 1.1. (p. 2) says something about the limit n,Nn , N \to \infty taken with fixed proportionality, namely such that nN 2n \propto N^2.

    In words, the statement of the theorem is roughly that in this limit the log of the Schur-Weyl probability density becomes “almost certainly approximately constant” , and that this constant scales as n\sqrt{n}, hence as NN – which should mean that in this limit the Shannon entropy of the Schur-Weyl measure is approximately unity times this constant nH c\sim \sqrt{n} \cdot H_c.

    But the Shannon entropy of the Schur-Weyl measure is a lower bound for our entropy.

    Not sure yet if this is good for anything, just saying.

    • CommentRowNumber10.
    • CommentAuthorUrs
    • CommentTimeJun 2nd 2021

    In particular, this result (mentioned in #9) seems to imply that in the limit nN 2n \propto N^2 \to \infty our Cayley-state probability distribution is

    (λ,i λ)e nH cdim(S (λ))=e nH c|sYTableaux λ|. (\lambda, i_\lambda) \;\mapsto\; \frac{ e^{\sqrt{n} \cdot H_c} }{dim\big(S^{(\lambda)}\big)} \;=\; \frac{ e^{\sqrt{n} \cdot H_c} }{ \left\vert sYTableaux_\lambda \right\vert } \,.

    This would reduce us to computing asymptotic numbers of standard Young tableaux without bound on their height, which is a question with some classical results to it…

    • CommentRowNumber11.
    • CommentAuthorDavid_Corfield
    • CommentTimeJun 3rd 2021

    That’s interesting. There’s a sign out, no?

    (λ,i λ)e nH c|sYTableaux λ|. (\lambda, i_\lambda) \;\mapsto\; \frac{ e^{-\sqrt{n} \cdot H_c} }{ \left\vert sYTableaux_\lambda \right\vert } \,.

    So we have the hook content formula for |sYTableaux λ|\left\vert sYTableaux_\lambda \right\vert.

    • CommentRowNumber12.
    • CommentAuthorUrs
    • CommentTimeJun 3rd 2021
    • (edited Jun 3rd 2021)

    Moreover, if we assume that the Schur-Weyl- and Plancherel-measures are asymptotically indistinguishable (as suggested on p. 3 here, though the way in which the double scaling limit is taken may matter) then we could use in the above formula the asymptotic estimate

    ln(|sYTableaux λ|)c2n12ln(n!) ln \big( \left\vert sYTableaux_\lambda \right\vert \big) \;\sim\; \frac{c}{2}\sqrt{n} - \tfrac{1}{2}ln(n!)

    (from here). Naively plugging in these asymptotic estimates into the formula for the Shannon entropy would then yield the following asymptotic expression for the entropy of the Cayley state:

    λPart(n)1i λ|sYTableaux λ|p Cay(λ,i λ)lnp Cay(λ,i λ) =λPart(n)p SW(λ)lne Hn|sYoungTableaux λ| n=cN 2λPart(n)e nH((Hc2)n+12ln(n!)) e n(H+π2/3)((Hc2)n+12ln(n!)) \begin{aligned} - \underset{ \mathclap{ {\lambda \in Part(n)} \atop { 1 \leq i_\lambda \leq \left\vert sYTableaux_\lambda \right\vert } } }{\sum} p^{Cay}(\lambda,i_\lambda) \cdot ln p^{Cay}(\lambda,i_\lambda) & = - \underset{ \mathclap{ {\lambda \in Part(n)} } }{\sum} p^{SW}(\lambda) \cdot ln \frac{e^{- H\cdot \sqrt{n}}}{ \left\vert sYoungTableaux_\lambda\right\vert } \\ & \underset{n = c N^2 \to \infty}{\sim} \underset{\lambda \in Part(n)}{\sum} e^{- \sqrt{n}\cdot H} \big( (- H - \tfrac{c}{2}) \cdot \sqrt{n} + \tfrac{1}{2} ln(n!) \big) \\ & \;\sim\; e^{\sqrt{n}\cdot (- H + \pi \sqrt{2/3} )} \big( (- H - \tfrac{c}{2}) \cdot \sqrt{n} + \tfrac{1}{2} ln(n!) \big) \end{aligned}

    In the last step I used the Hardy-Ramanujan asymptotics for |Part(n)|π2n/3\left\vert Part(n)\right\vert \sim \pi \sqrt{2n/3 } (taken from MO:103070 for the moment).

    Not claiming that this is correct, as there is handwaving involved in putting these different asymptotics together this way. But maybe it helps to think along these lines.

    • CommentRowNumber13.
    • CommentAuthorUrs
    • CommentTimeJun 3rd 2021

    Oh, I only now see #11. Yes, let me check the sign.

    • CommentRowNumber14.
    • CommentAuthorUrs
    • CommentTimeJun 3rd 2021

    By the way, a joint scaling limit of just this form

    n=cN 2 n = c N^2 \to \infty

    underlies the AdS-CFT correspondence: there

    N=λ(1g) 2 N = \lambda \left(\tfrac{1}{g}\right)^2 \to \infty

    with λ\lambda the ’t Hooft coupling.

    • CommentRowNumber15.
    • CommentAuthorDavid_Corfield
    • CommentTimeJun 3rd 2021

    Unfortunately I’ve got to spend today composing a reply to reviewers of a grant bid, with pretty poor prospects of success. This looks much more interesting.

    • CommentRowNumber16.
    • CommentAuthorUrs
    • CommentTimeJun 3rd 2021

    Sure, no rush. And I need to get back to equivariant bundles.

    BTW, I see now that the real asymptotic expansion of the partition function has a 1/n1/n factor (now here).

    With that included, the expansion of the Cayley entropy in #12 is of the form

    e an(b+cn 1/2+dn 1/2ln(n)). e^{ - a \cdot \sqrt{n} } \big( b + c \cdot n^{-1/2} + d \cdot n^{- 1/2} ln(n) \big) \,.

    If here we use #14 to identify ng 2n \propto g^{-2} and think of gg as a coupling constant, then this has the form of a non-perturbative contribution to a trans-series expansion of a Feynman series (here):

    e α/g(β+γg+δgln(g)) e^{- \alpha /g} \big( \beta + \gamma \cdot g + \delta \cdot g ln(g) \big)
    • CommentRowNumber17.
    • CommentAuthorDavid_Corfield
    • CommentTimeJun 3rd 2021

    if we assume that the Schur-Weyl- and Plancherel-measures are asymptotically indistinguishable

    I added a couple of papers to Pierre-Loïc Méliot which are relevant. Will need more time to digest.

    • CommentRowNumber18.
    • CommentAuthorDavid_Corfield
    • CommentTimeJun 4th 2021

    At the bottom of page 2 of

    there is discussion of the relationship between Schur-Weyl- and Plancherel-measures.

    • CommentRowNumber19.
    • CommentAuthorUrs
    • CommentTimeJun 4th 2021

    Could you check if you are pointing to the intended paper? On relation between the two measures this one seems to just mention that both have the same limit shape of Young diagrams for c=0c = 0. (Incidentally, from p. 5 in Meliot 11 I take it that they don’t quite have the same limit shape for c0c \neq 0, but the statement there remains vague.)

    I had been going by p. 3 of Mkrtchyan 11, which claims that at fixed nn the Schur-Weyl measure converges pointwise to the Plancherel measure as NN \to \infty, with pointer, “for example”, to Olshanski 09 – I had dug that out but still can’t read it, since it is in Russian. But it would be necessary to see the details, since we would need such a convergence to hold not for fixed nn, but for nN 2n \propto N^2 \to \infty.

    On the other hand, for the argument in #12 to work we’d need less than the two measures to converge – It would be sufficient that they have the same null sets in the limit.

    • CommentRowNumber20.
    • CommentAuthorDavid_Corfield
    • CommentTimeJun 4th 2021

    Sorry I got the wrong one. It’s bottom of p. 2 of

    • CommentRowNumber21.
    • CommentAuthorUrs
    • CommentTimeJun 4th 2021

    Thanks. So that’s indeed the kind of discussion as on that p. 5 in Meliot 11 that I had seen.

    I suppose the fact that the limit shapes for c>0c \gt 0 are different (though close, apparently) for the two measures means that the measures themselves can’t quite converge in the double scaling limit n=cN\sqrt{n} = c N \to \infty. Maybe they could still be close enough to have the same asymptotic null sets?

    By the way, I was vaguely wondering if those limit shapes are the answer to our quest for the maximum of the Schur-Weyl distribution: We had seen that the naive idea that the maximum is at λ=(n)\lambda = (n) receives some correction. But those limit shapes could possibly be thought of as a deformation of λ=(n)\lambda = (n) by adding a small(er) amount of boxes in further rows. This intuition is supported by the graphics on the very top of p. 4 of the article you point to, for what it’s worth.

    So I am wondering: Does the Schur-Weyl distribution in the limit n=cN\sqrt{n} = c N \to \infty take its maximum on that “deformed LSKV” limit shape? Is that maybe implied by those “central limit theorems” that these authors discuss, saying that the Schur-Weyl measure is a Gaussian fluctuation around that limit shape, of sorts (if I am understanding that right)?

Add your comments
  • Please log in or leave your comment as a "guest post". If commenting as a "guest", please include your name in the message as a courtesy. Note: only certain categories allow guest posts.
  • To produce a hyperlink to an nLab entry, simply put double square brackets around its name, e.g. [[category]]. To use (La)TeX mathematics in your post, make sure Markdown+Itex is selected below and put your mathematics between dollar signs as usual. Only a subset of the usual TeX math commands are accepted: see here for a list.

  • (Help)