Maximum Likelihood, permutohedra and Associativity Equations

No\'emie C. Combe

REVIEW 4 major objections 4 minor 41 references

Maximum Likelihood, permutohedra and Associativity Equations

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The interior of a diagonal spectrahedron satisfies the Associativity Equations, and ML degree is governed by torus-fixed boundary points.

desk verdict Interesting research program on spectrahedra and Frobenius manifolds, but the main proofs don't hold together. read the letter →

arxiv 2501.01345 v1 pith:4Q5EYYQW submitted 2025-01-02 math.AG math.DG

Noémie C. Combe This is my paper · ORCID

classification math.AGmath.DG MSC 14M2553D4562H0514N10

keywords maximumlikelihooddegreeFrobeniusmanifoldsAssociativityEquationspermutohedronspectrahedronWishartdistributionsconcentrationmatricestoricvarieties

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the cone of concentration matrices for linear Gaussian models carries a special geometric structure: it is a Monge–Ampère domain, and the log-likelihood function supplies its potential. On the smaller moduli space of diagonal matrices that parameterizes a polyhedral spectrahedron, the paper claims the Associativity Equations (WDVV) hold, so this optimization-theoretic object is a Frobenius manifold. Finally, it claims the maximum likelihood degree — the number of complex critical points of the likelihood equations — is indexed by torus-fixed points on a permutohedral boundary, called Frobenius residuals. If true, this connects algebraic statistics, semidefinite programming, and mirror symmetry in a concrete way.

What carries the argument

The diagospectrahedron, a spectrahedron parameterized by diagonal matrices, with potential $\Phi = \log \det$; the Hessian metric $g_{ij} = \partial_i \partial_j \Phi$ and its inverse, and the third-derivative tensors $A_{ijk} = \partial_i \partial_j \partial_k \Phi$. The Associativity (WDVV) equations for $\Phi$ are the equality $\sum_{e,f} g^{ef} A_{abe} A_{fcd} = \sum_{e,f} g^{ef} A_{ade} A_{fcb}$. The compactification is analyzed through Białynicki–Birula (BB) cells, affine-cell decompositions of a projective variety induced by a torus action, on the variety of complete quadrics; the torus-fixed points are vertices of a permutohedron, and these cells pave the boundary and define the Frobenius residuals.

What would settle it

For the $n=3$ diagonal spectrahedron, compute $\Phi=\log\det$ in the flat coordinates, form $g_{ij}$ and $A_{ijk}$, and check directly whether $\sum_{e,f} g^{ef} A_{abe} A_{fcd} = \sum_{e,f} g^{ef} A_{ade} A_{fcb}$ holds at a generic diagonal matrix; if not, the implication in Theorem 3 fails. Alternatively, compare the Theorem 4 sum over torus-fixed points on Frobenius residuals with a direct solving of the likelihood equations for a $3\times 3$ diagonal linear concentration model.

Watch

Extended reading notes

Core claim

The central discovery is that the geometry of the cone of concentration matrices, and in particular the diagonal spectrahedron inside it, obeys equations from 2D topological field theory. The paper proves the cone $S^n_{>0}$ is an elliptic Monge–Ampère domain (Theorem 2), that the log-likelihood function generates the potential function at the identity (Lemma 3), and that the interior of the diagospectrahedron is parametrized by a space satisfying the Associativity Equations (Theorem 3). It further proves that the compactification of the moduli space of diagonal matrices is a toric variety carrying a permutohedron structure, and that the ML degree is a sum over rational numbers indexed by $T$-fixed points lying on so-called Frobenius residuals (Theorem 4).

Load-bearing premise

The proof that the diagonal spectrahedron satisfies the Associativity Equations assumes that vanishing of the Riemann curvature of the Hessian metric alone implies the equality of third-derivative contractions in the WDVV equations, without constructing the deformed flat connection that standard Frobenius manifold theory requires.

Editorial extensions

If this is right

The feasible regions of semidefinite programs defined by diagonal spectrahedra are Frobenius manifolds, so numerical and algebraic methods from Frobenius manifold theory could be applied to optimization problems.
The ML degree of linear concentration models can be computed combinatorially from torus-fixed points on the permutohedral boundary, without solving the likelihood equations directly.
The compactified diagonal moduli space is a toric variety with permutohedron combinatorics, making it accessible to toric geometry and polytope algorithms.
The log-likelihood generating the potential at the identity gives a direct link between maximum likelihood estimation and Hessian/Kähler geometry of the cone.
The Monge–Ampère domain structure implies the cone is a viable setting for optimal transport and interpolation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper leaves implicit that if the WDVV equations hold on the whole diagospectrahedron, the deformed flat connection whose flatness is normally required for a Frobenius manifold should be explicitly constructible from $\Phi$; building it would strengthen Theorem 3.
One testable extension is to compute the rational numbers assigned to each torus-fixed point for small $n$ (say $n=3,4$) and compare their sum with directly computed ML degrees of diagonal Gaussian models.
The permutohedral boundary suggests a tropical limiting model for ML estimation, where degenerate concentration matrices on the boundary act as points at infinity whose contributions to the degree can be tracked combinatorially.
The analogy with Losev–Manin spaces hints at a factorization or recursion formula for ML degree across Frobenius residuals, mirroring recursion in quantum cohomology.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper's ambition is to tie diagospectrahedra to Frobenius manifolds and ML degree, and the questions are worth asking. The standard background—Hessian structure on symmetric cones, pre-Lie algebra from affine flatness, the permutohedral compactification of diagonal quadrics—is correctly assembled, and the survey of complete quadrics and BB cells is a nice recap for outsiders. The genuinely new pieces are the WDVV claim for diagonal spectrahedra and the Frobenius residuals, and I don't see those in the cited literature.

That's where the trouble starts. Theorem 3's proof uses Eq. (6) as the Riemann curvature tensor of a Hessian metric, but that formula is not the curvature. In affine coordinates the Levi-Civita curvature contains derivatives of the Christoffel symbols (fourth derivatives of the potential), not just the products of third derivatives with g^{-1}. So vanishing of the metric curvature gives no reason for the product c = g^{-1} A to be associative, and Eq. (7) does not follow. Without associativity, calling the moduli space a Frobenius manifold is unfounded. This is the load-bearing step, and it's wrong.

Lemma 3 is softer but similarly unsatisfying: putting K = Id in the log-likelihood gives exp(ℓ) = exp(-tr(S)), and the potential at the identity is defined by the integral of exp(-tr(S)). That's a definitional observation, not a generation result. Theorem 4 is even weaker: the assertion that the torus fixed points of complete quadrics 'happen to lie' on the Frobenius residuals is stated without argument, and the proof gives no mechanism.

The paper reads more like a research proposal with conjectures than a proof preprint. The standard parts are fine, the new parts are not established. I can't recommend citing it for the new results yet, and the self-citation network doesn't help.

That said, the questions are real and the author clearly has a program in mind. I'd send this to a serious referee if only to get a detailed account of why the Hessian geometry needs more work; the referee report could help the author reshape it. But as it stands, the paper should not be accepted. If the author can fix the curvature argument or honestly recast the claims as conjectures, it might become a useful note for the algebraic statistics and information geometry crowd. Right now it's a maybe for a reading group, but not something I'd build on.

Referee Report

4 major / 4 minor

Summary. The paper studies the cone of positive definite symmetric matrices (concentration matrices) and claims several structural results: that the cone is a Monge–Ampère domain (Theorem 2), that the log-likelihood generates the potential at the identity (Lemma 3), that the tangent sheaf carries a pre-Lie algebra structure, and that the interior of the diagospectrahedron satisfies the Associativity Equations (Theorem 3). It then introduces "Frobenius residuals" via the BB-cell decomposition of a compactified toric diagonal space and claims that the maximum likelihood degree is a sum of rational numbers indexed by torus-fixed points lying on those residuals (Theorem 4). The paper is written as a sequence of short, mostly asserted statements with few detailed derivations.

Significance. If the central theorems were established, the paper would forge an appealing connection between algebraic statistics, Frobenius manifold theory, and permutohedral combinatorics: diagonal linear Gaussian models would form a Frobenius manifold, and the ML degree would be governed by toric boundary data. The claimed links to Losev–Manin spaces and to mirror symmetry would also be of interest. However, the present text does not provide the necessary mathematical support for these claims. The proof of Theorem 3 rests on an unproved and nonstandard curvature formula, Theorem 4 is asserted rather than proved, and Theorem 2 invokes a bounded-domain theorem for an unbounded cone. The paper contains no machine-checked proofs, no reproducible code, and no numerical examples that would compensate for these gaps.

major comments (4)

[Section 2, Theorem 3, Eq. (6)] Equation (6) is not the Riemann curvature tensor of a Hessian metric in affine coordinates. For a Hessian metric g_{ij}=∂_i∂_jΦ with Christoffel symbols Γ^k_{ij}=1/2 g^{kl}A_{ijl}, the Levi-Civita curvature contains derivatives of Γ, hence fourth derivatives of Φ, together with quadratic terms in Γ. The displayed expression contains only products of third derivatives contracted with g^{-1}, so it cannot be the full curvature tensor. Consequently, vanishing of the metric curvature gives no reason for the product c^i_{jk}=g^{il}A_{ljk} to be associative, which is exactly what Eq. (7) asserts. The proof must either justify Eq. (6) from a correct curvature formula under additional hypotheses, or construct the deformed flat connection and prove associativity as required in Frobenius manifold theory. As written, Theorem 3 is not proved.
[Section 4.4, Theorem 4] Theorem 4 is not derived. The proof consists of the single assertion that the torus-fixed points of [MMW2021] "happen to lie" on the Frobenius residuals. This is a substantive incidence statement: the ML-degree formula of [MMW2021] is indexed by torus-fixed points on the variety of complete quadrics, and the paper must identify those fixed points with vertices of the permutohedron and prove their containment in the specific boundary set Z\Z defined in Definition 2. The definition of Frobenius residuals as elements of Z\Z also does not, by itself, imply that all or even some of the relevant T-fixed points lie on them. Without this identification, Theorem 4 is only a conjecture.
[Section 3.2, Lemma 3] Lemma 3 is effectively a definitional identity. The proof observes that exp ℓ_S(Id)=exp{-tr(S)} and that ln χ(Id)=ln ∫_{Ω*} exp{-tr(S)}dS, which is precisely the integral used to define the characteristic function. Thus the statement "the log-likelihood generates the potential at the identity" reduces to a reformulation of the definition of χ. The lemma does not show that ℓ_S(K) for K≠Id is related to the potential, nor does it explain how ℓ generates the potential on the whole cone K_L. The claimed generation property therefore lacks the load-bearing content needed for the subsequent use of ℓ in the Frobenius-type structure.
[Section 2.1.5, Theorem 2 and Corollary 1] The proof of Theorem 2 invokes the classical Dirichlet problem for elliptic Monge–Ampère equations, but the cited theorem of Rauch–Taylor applies to bounded strictly convex domains, whereas S^n_{>0} is unbounded and not strictly convex. The existence of a smooth convex function Φ satisfying det Hess Φ = f on the whole cone is not established. Corollary 1 then asserts that the interior of the diagospectrahedron is parametrized by a flat elliptic Monge–Ampère subdomain, based on the sentence "a computation shows that the sectional curvature vanishes," but no computation or reference is provided. Since Corollary 1 is used as input to Theorem 3, these gaps affect the central claim of the paper.

minor comments (4)

[Definition 2 and Theorem 4] Definition 2 calls Frobenius residuals elements of Z\Z, while Theorem 4 speaks of connected components lying on the residuals; the terminology should be made consistent.
[Corollary 1] The phrase "a computation shows that the sectional curvature vanishes" should be replaced by an explicit curvature calculation or a precise reference, since the flatness assertion is used later.
[Section 4.3] The correspondence between torus-fixed points, BB-cell centers, and vertices of the permutohedron is asserted without proof; a clear statement of the bijection and its source would help the reader.
[General] There are several typographical and referencing issues: the citation [KZ] lacks full publication details, the name "Bia lynicki" appears with inconsistent spacing, and the use of n both for the matrix size and for the dimension of the cone in Section 2.1.3 is confusing.

Circularity Check

3 steps flagged · score 8.0 of 10

Central Theorem 3 is circular: Eq. (6) defines 'curvature' as the WDVV associator, so its vanishing is the Associativity Equations; Lemma 3 and Theorem 4 are definitional/renaming.

self definitional [Theorem 3 proof, Eq. (6)-(7), section on Associativity Equations]
"The Riemann curvature tensor can be expressed in terms of the rank three symmetric tensors as follows: (6) R_acdb = Σ_{e,f} g^{ef}(A_{eab}A_{fcd} − A_{ead}A_{fcb}). By the flatness hypothesis on spectrahedron, the curvature tensor vanishes. Therefore, from Eq.6 it follows that we obtain the equality Σ_{e,f} g^{ef} A_{eab}A_{fcd} = Σ_{e,f} g^{ef} A_{ead}A_{fcb}. ... Therefore, it follows that on the spectrahedron the following holds: (7) ∀a,b,c,d: Σ ∂_a∂_b∂_e Φ g^{ef} ∂_f∂_c∂_d Φ = Σ ∂_f∂_c∂_b Φ g^{ef} ∂_e∂_a∂_d Φ."

The right-hand side of Eq. (6) is precisely the associator of the product c^i_{jk} = g^{il} A_{ljk}: it is the difference c^e_{ab} c_{ecd} − c^e_{ad} c_{ecb}. Setting R = 0 is exactly the associativity/WDVV equation (7). Thus the proof does not derive associativity from a separately established flatness; it defines the 'curvature' to be the associator and then lets 'flatness' carry the conclusion. No theorem is cited showing that the Levi-Civita curvature of the Hessian metric g_{ij}=∂_i∂_jΦ equals this expression; indeed that curvature contains fourth derivatives of Φ, so the equation is not the standard curvature formula. The conclusion is contained in the defining equality.
self definitional [Section 3.2, Lemma 3 and its proof]
"So, using the log-likelihood function, we can define the potential function at the identity by: ln χ(Id) = ln ∫_{Ω∗} exp{−⟨Id, S⟩}dS = ln ∫_{Ω∗} exp{−Tr(Id·S)}dS."

χ(Id) was already defined in Eq. (3) as ∫_{Ω∗} exp{−⟨x,a∗⟩} da∗ evaluated at x = Id; and ℓ_S(K)=log det K − tr(KS) gives exp ℓ_S(Id)=exp(−tr S)=exp(−⟨Id,S⟩). So 'the log-likelihood generates the potential at the identity' is a restatement of the definition of χ, not a generation theorem. The integrand of the characteristic function at the identity is, by construction, exp of the log-likelihood at the identity.

1 more flagged steps

renaming known result [Section 4.4, Theorem 4 proof]
"By [MMW2021], the M L-degree can be expressed as a sum, over a finite number of torus fixed points on the variety of complete quadrics Q, of rational numbers assigned to each of them. The torus fixed points coincide with the vertices of a permutohedron. These torus fixed points happen to lie on the connected components of the Frobenius residuals."

The theorem adds no computation; it imports the MMW2021 sum and reindexes it by 'Frobenius residuals'. Since Frobenius residuals are defined (Definition 2) as the complement of the Frobenius domain in its compactification, the torus fixed points of the permutohedral compactification are boundary points; the claim that they 'happen to lie' on residuals is an assertion identifying the new name with the old index set. The ML-degree statement therefore inherits its content from MMW2021 and is repackaged.

full rationale

The paper's main theorem (Theorem 3) is not self-contained: its proof declares the associator to be the Riemann curvature tensor (Eq. 6) and then infers WDVV from 'flatness' (Eq. 7). Since Eq. 6's right-hand side is exactly the difference whose vanishing is Eq. 7, the conclusion is built into the definition; this is a definitional circularity, not a derived result. Lemma 3 similarly restates the definition of χ(Id) in terms of exp ℓ_S(Id). Theorem 4 imports the ML-degree sum from [MMW2021] and only renames the summation index set as 'Frobenius residuals' by asserting the fixed points lie there. Other parts of the paper (Monge-Ampère domain, pre-Lie tangent sheaf, permutohedral toric compactification) are independent and not circular; however the central 'Frobenius manifold / ML-degree' claims are forced by definitional identifications. Self-citations alone are not the issue; [MMW2021] is external. Score 8: the central result reduces to a defining identity.

Assumptions & free parameters 0 free parameters · 6 assumptions · 1 invented entities

The paper introduces no fitted parameters, uses standard symmetric-cone facts as background, and postulates several unproved implications for the new claims. The main invented object, Frobenius residuals, has no independent evidence.

assumptions (6)

domain assumption The standard Monge-Ampere Dirichlet theory on bounded strictly convex domains applies to the unbounded cone S^n_{>0}.
Invoked in Theorem 2 proof using [RT77]; no boundedness or boundary data is discussed.
ad hoc to paper Vanishing of the Riemann curvature of the Hessian metric on a flat affine domain implies the WDVV Associativity Equations.
Used in Theorem 3 proof to pass from Eq. (6) to Eq. (7); no supporting theorem or construction of the deformed connection is given.
ad hoc to paper The compactified space of diagonal matrices is a toric variety whose T-fixed points are all vertices of the permutohedron and lie on Frobenius residuals.
Underpins Proposition 3 and Theorem 4; the containment on residuals is asserted without proof.
ad hoc to paper The ML-degree formula of [MMW2021] can be restricted to a sum over T-fixed points lying on Frobenius residuals.
Theorem 4 proof relies on this restriction; the cited paper only gives a sum over torus fixed points on complete quadrics.
standard math Pre-Lie algebra on the tangent sheaf follows from the affine flat structure of the cone.
Lemma 1 proof; this is a standard Koszul-Nomizu framework, but its application to this cone is only sketched.
standard math Symmetric cones carry a canonical Hessian metric from the characteristic function.
Lemma 2 from [FK94]; standard, though the sign written as g=-Hess ln chi conflicts with positive definiteness.

invented entities (1)

Frobenius residuals
purpose: Boundary points or components of the compactified Frobenius manifold of diagonal matrices, intended to index ML degree contributions.
No independent falsifiable prediction is provided; their existence and permutohedral structure are asserted from BB cells and 'happen to lie' statements.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Maximum Likelihood, permutohedra and Associativity Equations." pith.science (2026). https://pith.science/paper/4Q5EYYQW

@misc{pith2026250101345,
  author       = {Pith},
  title        = {Pith review of: Maximum Likelihood, permutohedra and Associativity Equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4Q5EYYQW}},
  note         = {Machine review of arXiv:2501.01345}
}

read the original abstract

We consider the cone of concentration matrices related to linear concentration models and Wishart laws. We prove that this cone is a Monge--Amp\`ere domain and that the log-likelihood function generates its potential function at the identity. The tangent sheaf carries the structure of a pre-Lie algebra. We also show that the moduli space of diagonal matrices parameterizing the polyhedral spectrahedron satisfies the Associativity Equations, a notion central in mirror symmetry, and that its compactification is a toric variety associated to a permutohedron, reminiscent to Losev--Manin spaces. Finally we introduce Frobenius residuals: these are connected components of the compactified Frobenius manifold of diagonal matrices, generated by the Bia\l{}ynicki--Birula cells. We prove that the Maximum Likelihood degree is indexed by components lying on those Frobenius residuals.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

[1]

Andersson, G

S. Andersson, G. Wojnar Wishart Distributions on Homogeneous Cones, Journal of Theoretical Probability 17 (2004) 781--818

work page 2004
[2]

Bia ynicki-Birula Some theorems on actions of algebraic groups

A. Bia ynicki-Birula Some theorems on actions of algebraic groups. Ann. Math. 2 (98), (1973), 480–497

work page 1973
[3]

Casalis, Les families exponentielles \`a variance quadratique homog\`ene sont de lois de Wishart sur un cone symétrique

M. Casalis, Les families exponentielles \`a variance quadratique homog\`ene sont de lois de Wishart sur un cone symétrique . C.R. Acad. Sci. Paris Ser. I Math. 312 , (1991), 537–540

work page 1991
[4]

Casalis, G

M. Casalis, G. Letac, The Lukacs-Olkin-Rubin characterization of the Wishart distribution on symmetric cones. Ann. Stat. 24, (1996), 763–786

work page 1996
[5]

Chentsov, Geometry of the manifold of probability distributions , Dokl

N. Chentsov, Geometry of the manifold of probability distributions , Dokl. Akad. Nauk SSSR 158:3 (1964) 543--546

work page 1964
[6]

Chentsov, Categories of mathematical statistics, Dokl

N. Chentsov, Categories of mathematical statistics, Dokl. Akad. Nauk SSSR 164:3 (1965) 511--514

work page 1965
[7]

N. C. Combe, On Frobenius structures in symmetric cones arXiv:2309.04334

work page Pith review arXiv
[8]

N.C: Combe, Landau-Ginzburg models, Monge-Ampere domains and (pre-)Frobenius manifolds arXiv:2409.00835

work page Pith review arXiv

Show all 41 references

[9]

N. C. Combe, On the geometry of Kähler--Frobenius manifolds and their classification arXiv:2411.14362

arXiv
[10]

N. C. Combe, Learning on hexagonal structures and Monge-Ampère operators arXiv:2412.04407

arXiv
[11]

N. C. Combe, Wishart cones and quantum geometry arXiv:2412.12289

arXiv
[12]

N. C. Combe, P. Combe, H. Nencka, Statistical manifolds and geometric invariants. Geometric Science of Information (2020) 565--573, Editors: F. Nielsen, F. Barbaresco

2020
[13]

N. C. Combe, P. Combe, H. Nencka, Algebraic properties of the information geometry's fourth Frobenius manifold. in Lecture Notes in Networks and Systems Springer series, (2022) 356-370

2022
[14]

N. C. Combe, Yu. Manin, F--manifolds and geometry of information, Bull. LMS, 52 , Issue 5 (2020), 777-792

2020
[15]

Combe, Y

N.C. Combe, Y. I. Manin, M. Marcolli, Moufang patterns and geometry of information , Pure and Applied Mathematics Quarterly 19 , Number 1, (2023) 149--189

2023
[16]

Combe, Y

N.C. Combe, Y. I. Manin, M. Marcolli, Geometry of Information: classical and quantum aspects , Journal of Theoretical Computer Sciences 908 (2022) 2–27

2022
[17]

DeConcini, M

C. DeConcini, M. Goresky, R. MacPherson, C. Procesi, On the geometry of quadrics and their degenerations. Commentarii Mathematici Helvetici, 63(1), (1988), 337–413

1988
[18]

J. A. Finat, A combinatorial presentation of the variety of complete quadrics. , Geometrie algebrique et applications, I (1984)

1984
[19]

Faraut, A

J. Faraut, A. Koranyi, Analysis on symmetric cones , Oxford Mathematical Monographs (1994)

1994
[20]

Gårding, An inequality for hyperbolic polynomials , J

L. Gårding, An inequality for hyperbolic polynomials , J. Math. Mech. 8 no. 2 (1959), 957-965

1959
[21]

Helgason, Differential geometry, Lie groups and Symmetric spaces , New York: Academic Press; (1978)

S. Helgason, Differential geometry, Lie groups and Symmetric spaces , New York: Academic Press; (1978)

1978
[22]

G. Khan J. Zhang, Statistical mirror symmetry Differential Geometry and its Applications 73 (2020)

2020
[23]

Koszul, Domaines born\'es homog\`enes et orbites des groupes de transformations affines, Bull

J-L. Koszul, Domaines born\'es homog\`enes et orbites des groupes de transformations affines, Bull. Soc. Math France 89 (1961) 515-533

1961
[24]

Losev, Yu

A. Losev, Yu. Manin, New Moduli Spaces of Pointed Curves and Pencils of Flat Connections , Michigan Math. J. 48 (2000), 443--472

2000
[25]

ojasiewicz, Triangulation of semi-analytic sets, Annali della Scuola Normale Superiore di Pisa - Classe di Scienze 18.4 (1964): 449-474

S. ojasiewicz, Triangulation of semi-analytic sets, Annali della Scuola Normale Superiore di Pisa - Classe di Scienze 18.4 (1964): 449-474

1964
[26]

Ronco, Permutads , Journal of Combinatorial Theory, Series A, Volume 120, Issue 2, (2013), Pages 340-365,

J-L Loday, M. Ronco, Permutads , Journal of Combinatorial Theory, Series A, Volume 120, Issue 2, (2013), Pages 340-365,

2013
[27]

Yu. I. Manin, Three constructions of Frobenius manifolds: a comparative study, Asian J. Math. 3:1 (1999) (Atiyah's Festschrift), 179--220

1999
[28]

Yu. I. Manin, Frobenius Manifolds, Quantum Cohomology, and Moduli Spaces , AMS Colloquium Publications, 47 (1999)

1999
[29]

Yu. I. Manin, Manifolds with multiplication on the tangent sheaf , Rend. Mat. Appl. 7 , 26:1 (2006), 69–85

2006
[30]

Micha ek, L

M. Micha ek, L. Monin and J. Wi\' s niewski, Maximum Likelihood Degree, Complete Quadrics, and C ^* -Action , SIAM Journal on Applied Algebra and Geometry 5 , 1 , (2021), 60-85

2021
[31]

Manivel, M

L. Manivel, M. Micha ek, L. Monin, T. Seynnaeve, M. Vodi cka, Complete quadrics: Schubert calculus for Gaussian models and semidefinite programming J. Eur. Math. Soc. 26, 3091–3135 (2024) European Mathematical Society

2024
[32]

Nomizu, Invariant affine connections on homogeneous spaces , Amer

K. Nomizu, Invariant affine connections on homogeneous spaces , Amer. J. Math. 76 (1) (1954) 33--65

1954
[33]

Pennec, 3-Manifold-valued image processing with SPD matrices Riemannian Geometric Statistics in Medical Image Analysis (2020), 75-134

X. Pennec, 3-Manifold-valued image processing with SPD matrices Riemannian Geometric Statistics in Medical Image Analysis (2020), 75-134

2020
[34]

Ramana, A

M. Ramana, A. J. Goldman, Some geometric results in semidefinite programming. Journal of Global Optimization, 7(1), (1995) 33–50

1995
[35]

J. B. Rauch, B. A. Taylor, The Dirichlet problem for the multidimensional Monge- Amp\`ere equation , Rocky Mountain J. Math. 7 (1977) 345-364

1977
[36]

Sturmfels, C

B. Sturmfels, C. Uhler, Multivariate Gaussians, semidefinite matrix completion, and convex algebraic geometry. Ann Inst Stat Math 62, (2010) 603–638

2010
[37]

Thaddeus, Complete collineations revisited

M. Thaddeus, Complete collineations revisited. Math. Ann. 315(3), (1999) 469–495

1999
[38]

Totaro, The curvature of a Hessian metric, International Journal of Mathematics, World Scientific Publishing Company Vol

B. Totaro, The curvature of a Hessian metric, International Journal of Mathematics, World Scientific Publishing Company Vol. 15, No. 4 (2004) 369--391

2004
[39]

Vainsencher, Complete collineations and blowing up determinantal ideals, Mathematische Annalen, 267(3), (1984) 417–432

I. Vainsencher, Complete collineations and blowing up determinantal ideals, Mathematische Annalen, 267(3), (1984) 417–432

1984
[40]

P. M. H. Wilson, Sectional curvatures of K\"ahler moduli , Math. Ann. 330 (2004) 631--664

2004
[41]

Wishart, The generalized product moment distribution in samples from a normal multivariate population, Biometrika, 20A (1928) 32-52

J. Wishart, The generalized product moment distribution in samples from a normal multivariate population, Biometrika, 20A (1928) 32-52

1928

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Andersson, G

S. Andersson, G. Wojnar Wishart Distributions on Homogeneous Cones, Journal of Theoretical Probability 17 (2004) 781--818

work page 2004

[2] [2]

Bia ynicki-Birula Some theorems on actions of algebraic groups

A. Bia ynicki-Birula Some theorems on actions of algebraic groups. Ann. Math. 2 (98), (1973), 480–497

work page 1973

[3] [3]

Casalis, Les families exponentielles \`a variance quadratique homog\`ene sont de lois de Wishart sur un cone symétrique

M. Casalis, Les families exponentielles \`a variance quadratique homog\`ene sont de lois de Wishart sur un cone symétrique . C.R. Acad. Sci. Paris Ser. I Math. 312 , (1991), 537–540

work page 1991

[4] [4]

Casalis, G

M. Casalis, G. Letac, The Lukacs-Olkin-Rubin characterization of the Wishart distribution on symmetric cones. Ann. Stat. 24, (1996), 763–786

work page 1996

[5] [5]

Chentsov, Geometry of the manifold of probability distributions , Dokl

N. Chentsov, Geometry of the manifold of probability distributions , Dokl. Akad. Nauk SSSR 158:3 (1964) 543--546

work page 1964

[6] [6]

Chentsov, Categories of mathematical statistics, Dokl

N. Chentsov, Categories of mathematical statistics, Dokl. Akad. Nauk SSSR 164:3 (1965) 511--514

work page 1965

[7] [7]

N. C. Combe, On Frobenius structures in symmetric cones arXiv:2309.04334

work page Pith review arXiv

[8] [8]

N.C: Combe, Landau-Ginzburg models, Monge-Ampere domains and (pre-)Frobenius manifolds arXiv:2409.00835

work page Pith review arXiv