random matrix
{{Short description|Matrix-valued random variable}}
In probability theory and mathematical physics, a random matrix is a matrix-valued random variable—that is, a matrix in which some or all of its entries are sampled randomly from a probability distribution. Random matrix theory (RMT) is the study of properties of random matrices, often as they become large. RMT provides techniques like mean-field theory, diagrammatic methods, the cavity method, or the replica method to compute quantities like traces, spectral densities, or scalar products between eigenvectors. Many physical phenomena, such as the spectrum of nuclei of heavy atoms,{{Cite journal |last=Wigner |first=Eugene P. |date=1955 |title=Characteristic Vectors of Bordered Matrices With Infinite Dimensions |url=https://www.jstor.org/stable/1970079 |journal=Annals of Mathematics |volume=62 |issue=3 |pages=548–564 |doi=10.2307/1970079 |jstor=1970079 |issn=0003-486X}}{{Cite report |url=https://www.osti.gov/biblio/4319287 |title=Conference on Neutron Physics by Time-Of-Flight Held at Gatlinburg, Tennessee, November 1 and 2, 1956 |editor-last=Block |editor-first=R. C. |editor-last2=Good |editor-first2=W. M. |date=1957-07-01 |publisher=Oak Ridge National Lab.|location=Oak Ridge, Tennessee |type=Report ORNL-2309 |doi=10.2172/4319287 |osti=4319287 |language=English |editor-last3=Harvey |editor-first3=J. A. |editor-last4=Schmitt |editor-first4=H. W. |editor-last5=Trammell |editor-first5=G. T.}} the thermal conductivity of a lattice, or the emergence of quantum chaos, can be modeled mathematically as problems concerning large, random matrices.
Applications
=Physics=
In nuclear physics, random matrices were introduced by Eugene Wigner to model the nuclei of heavy atoms. Wigner postulated that the spacings between the lines in the spectrum of a heavy atom nucleus should resemble the spacings between the eigenvalues of a random matrix, and should depend only on the symmetry class of the underlying evolution.{{harvnb|Mehta|2004}} In solid-state physics, random matrices model the behaviour of large disordered Hamiltonians in the mean-field approximation.
In quantum chaos, the Bohigas–Giannoni–Schmit (BGS) conjecture asserts that the spectral statistics of quantum systems whose classical counterparts exhibit chaotic behaviour are described by random matrix theory.{{cite journal| last1=Bohigas|first1=O.| last2=Giannoni|first2=M.J.| last3=Schmit|first3=Schmit| title=Characterization of Chaotic Quantum Spectra and Universality of Level Fluctuation Laws|journal=Phys. Rev. Lett.| year=1984|volume=52|issue=1| pages=1–4| doi=10.1103/PhysRevLett.52.1| bibcode=1984PhRvL..52....1B}}
In quantum optics, transformations described by random unitary matrices are crucial for demonstrating the advantage of quantum over classical computation (see, e.g., the boson sampling model).{{cite journal| last1=Aaronson|first1=Scott| last2=Arkhipov|first2=Alex| title=The computational complexity of linear optics |journal=Theory of Computing |date=2013 |volume=9 |pages=143–252 |doi=10.4086/toc.2013.v009a004 |doi-access=free}} Moreover, such random unitary transformations can be directly implemented in an optical circuit, by mapping their parameters to optical circuit components (that is beam splitters and phase shifters).{{cite journal |last1=Russell|first1=Nicholas |last2=Chakhmakhchyan|first2=Levon | last3=O'Brien|first3=Jeremy |last4=Laing|first4=Anthony |journal=New J. Phys. | title=Direct dialling of Haar random unitary matrices| date=2017| volume=19|issue=3| pages=033007| doi=10.1088/1367-2630/aa60ed |bibcode=2017NJPh...19c3007R |arxiv=1506.06220 | s2cid=46915633}}
Random matrix theory has also found applications to the chiral Dirac operator in quantum chromodynamics,{{cite journal |vauthors=Verbaarschot JJ, Wettig T |title=Random Matrix Theory and Chiral Symmetry in QCD|journal=Annu. Rev. Nucl. Part. Sci. |volume=50|pages=343–410 |year=2000 |doi=10.1146/annurev.nucl.50.1.343 |arxiv = hep-ph/0003017 |bibcode = 2000ARNPS..50..343V |s2cid=119470008}} quantum gravity in two dimensions,{{cite journal |vauthors=Franchini F, Kravtsov VE |title=Horizon in random matrix theory, the Hawking radiation, and flow of cold atoms |journal=Phys. Rev. Lett. |volume=103 |issue=16 |pages=166401|date=October 2009 |pmid=19905710|doi=10.1103/PhysRevLett.103.166401 |bibcode = 2009PhRvL.103p6401F |arxiv = 0905.3533 |s2cid=11122957 }} mesoscopic physics,{{cite journal |vauthors=Sánchez D, Büttiker M |title=Magnetic-field asymmetry of nonlinear mesoscopic transport |journal=Phys. Rev. Lett. |volume=93 |issue=10 |pages=106802 |date=September 2004 |pmid=15447435 |doi=10.1103/PhysRevLett.93.106802 |bibcode=2004PhRvL..93j6802S | arxiv = cond-mat/0404387 |s2cid=11686506 }} spin-transfer torque,{{cite journal |vauthors=Rychkov VS, Borlenghi S, Jaffres H, Fert A, Waintal X |title=Spin torque and waviness in magnetic multilayers: a bridge between Valet-Fert theory and quantum approaches |journal=Phys. Rev. Lett. |volume=103 |issue=6 |pages=066602 |date=August 2009 |pmid=19792592|doi=10.1103/PhysRevLett.103.066602|bibcode=2009PhRvL.103f6602R|arxiv = 0902.4360 |s2cid=209013 }} the fractional quantum Hall effect,{{cite journal |author=Callaway DJE |title=Random matrices, fractional statistics, and the quantum Hall effect |journal=Phys. Rev. B |volume=43 |issue=10 |pages=8641–8643 |date=April 1991 |pmid=9996505 |doi=10.1103/PhysRevB.43.8641 |bibcode = 1991PhRvB..43.8641C |author-link=David J E Callaway }} Anderson localization,{{cite journal |vauthors=Janssen M, Pracz K |title=Correlated random band matrices: localization-delocalization transitions |journal=Phys. Rev. E |volume=61 |issue=6 Pt A |pages=6278–86 |date=June 2000 |pmid=11088301 |doi=10.1103/PhysRevE.61.6278 |arxiv = cond-mat/9911467 |bibcode = 2000PhRvE..61.6278J |s2cid=34140447 }} quantum dots,{{cite journal |vauthors=Zumbühl DM, Miller JB, Marcus CM, Campman K, Gossard AC |title=Spin-orbit coupling, antilocalization, and parallel magnetic fields in quantum dots |journal=Phys. Rev. Lett. |volume=89 |issue=27 |pages=276803 |date=December 2002 |pmid=12513231 | doi=10.1103/PhysRevLett.89.276803 |bibcode=2002PhRvL..89A6803Z |arxiv = cond-mat/0208436 |s2cid=9344722 }} and superconductors{{cite journal |author=Bahcall SR |title=Random Matrix Model for Superconductors in a Magnetic Field |journal=Phys. Rev. Lett. |volume=77 |issue=26 |pages=5276–5279 |date=December 1996|pmid=10062760|doi=10.1103/PhysRevLett.77.5276|bibcode=1996PhRvL..77.5276B|arxiv = cond-mat/9611136 |s2cid=206326136 }}
=Mathematical statistics and numerical analysis=
In multivariate statistics, random matrices were introduced by John Wishart, who sought to estimate covariance matrices of large samples.{{harvnb|Wishart|1928}} Chernoff-, Bernstein-, and Hoeffding-type inequalities can typically be strengthened when applied to the maximal eigenvalue (i.e. the eigenvalue of largest magnitude) of a finite sum of random Hermitian matrices.{{cite journal | last=Tropp | first=J.|title=User-Friendly Tail Bounds for Sums of Random Matrices |journal=Foundations of Computational Mathematics |year=2011 |doi=10.1007/s10208-011-9099-z|volume=12 | issue=4| pages=389–434 |arxiv=1004.4389| s2cid=17735965}} Random matrix theory is used to study the spectral properties of random matrices—such as sample covariance matrices—which is of particular interest in high-dimensional statistics. Random matrix theory also saw applications in neural networks{{cite journal |last1=Pennington |first1=Jeffrey |last2=Bahri |first2=Yasaman |date=2017 |title=Geometry of Neural Network Loss Surfaces via Random Matrix Theory |journal=ICML'17: Proceedings of the 34th International Conference on Machine Learning |volume=70|s2cid=39515197 }} and deep learning, with recent work utilizing random matrices to show that hyper-parameter tunings can be cheaply transferred between large neural networks without the need for re-training.{{cite arXiv |last=Yang |first=Greg |date=2022 |title=Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer |class=cs.LG |eprint=2203.03466v2}}
In numerical analysis, random matrices have been used since the work of John von Neumann and Herman Goldstine{{harvnb|von Neumann|Goldstine|1947}} to describe computation errors in operations such as matrix multiplication. Although random entries are traditional "generic" inputs to an algorithm, the concentration of measure associated with random matrix distributions implies that random matrices will not test large portions of an algorithm's input space.{{harvnb|Edelman|Rao|2005}}
=Number theory=
In number theory, the distribution of zeros of the Riemann zeta function (and other L-functions) is modeled by the distribution of eigenvalues of certain random matrices.{{cite journal| last=Keating|first=Jon|title=The Riemann zeta-function and quantum chaology| journal=Proc. Internat. School of Phys. Enrico Fermi |year=1993 | volume=CXIX| pages=145–185| doi=10.1016/b978-0-444-81588-0.50008-0| isbn=9780444815880}} The connection was first discovered by Hugh Montgomery and Freeman Dyson. It is connected to the Hilbert–Pólya conjecture.
=Free probability=
The relation of free probability with random matricesMingo, James A.; Speicher, Roland (2017): [http://rolandspeicher.com/literature/mingo-speicher/ Free Probability and Random Matrices]. Fields Institute Monographs, Vol. 35, Springer, New York is a key reason for the wide use of free probability in other subjects. Voiculescu introduced the concept of freeness around 1983 in an operator algebraic context; at the beginning there was no relation at all with random matrices. This connection was only revealed later in 1991 by Voiculescu;Voiculescu, Dan (1991): "Limit laws for random matrices and free products". Inventiones mathematicae 104.1: 201-220 he was motivated by the fact that the limit distribution which he found in his free central limit theorem had appeared before in Wigner's semi-circle law in the random matrix context.
=Computational neuroscience=
In the field of computational neuroscience, random matrices are increasingly used to model the network of synaptic connections between neurons in the brain. Dynamical models of neuronal networks with random connectivity matrix were shown to exhibit a phase transition to chaos{{cite journal| last=Sompolinsky|first=H. |author2=Crisanti, A. |author3=Sommers, H. |title=Chaos in Random Neural Networks |journal=Physical Review Letters|date=July 1988|volume=61|issue=3|pages=259–262|doi=10.1103/PhysRevLett.61.259|pmid=10039285|bibcode = 1988PhRvL..61..259S |s2cid=16967637 }} when the variance of the synaptic weights crosses a critical value, at the limit of infinite system size. Results on random matrices have also shown that the dynamics of random-matrix models are insensitive to mean connection strength. Instead, the stability of fluctuations depends on connection strength variation{{cite journal|last=Rajan|first=Kanaka|author2=Abbott, L. |title=Eigenvalue Spectra of Random Matrices for Neural Networks| journal=Physical Review Letters| date=November 2006|volume=97|issue=18| pages=188104 |doi=10.1103/PhysRevLett.97.188104 | pmid=17155583 |bibcode = 2006PhRvL..97r8104R }}{{cite journal|last=Wainrib|first=Gilles|author2=Touboul, Jonathan |title=Topological and Dynamical Complexity of Random Neural Networks|journal=Physical Review Letters |date=March 2013|volume=110|issue=11 |doi=10.1103/PhysRevLett.110.118101 | arxiv = 1210.5082 |bibcode = 2013PhRvL.110k8101W |pmid=25166580|page=118101|s2cid=1188555}} and time to synchrony depends on network topology.{{cite journal |last=Timme|first=Marc |author2=Wolf, Fred |author3=Geisel, Theo |title=Topological Speed Limits to Network Synchronization | journal=Physical Review Letters| date=February 2004|volume=92 | issue=7| doi=10.1103/PhysRevLett.92.074101|arxiv = cond-mat/0306512 |bibcode = 2004PhRvL..92g4101T |pmid=14995853 |page=074101|s2cid=5765956}}{{cite journal |last1=Muir|first1=Dylan | last2=Mrsic-Flogel|first2=Thomas |title=Eigenspectrum bounds for semirandom matrices with modular and spatial structure for neural networks |journal=Phys. Rev. E|date=2015|volume=91 |issue=4 |page=042808 |doi=10.1103/PhysRevE.91.042808|pmid=25974548|bibcode = 2015PhRvE..91d2808M |url=http://edoc.unibas.ch/41441/1/20160120100936_569f4ed0ddeee.pdf}}
In the analysis of massive data such as fMRI, random matrix theory has been applied in order to perform dimension reduction. When applying an algorithm such as PCA, it is important to be able to select the number of significant components. The criteria for selecting components can be multiple (based on explained variance, Kaiser's method, eigenvalue, etc.). Random matrix theory in this content has its representative the Marchenko-Pastur distribution, which guarantees the theoretical high and low limits of the eigenvalues associated with a random variable covariance matrix. This matrix calculated in this way becomes the null hypothesis that allows one to find the eigenvalues (and their eigenvectors) that deviate from the theoretical random range. The components thus excluded become the reduced dimensional space (see examples in fMRI {{Cite journal |last1=Vergani |first1=Alberto A. | last2=Martinelli | first2=Samuele |last3=Binaghi |first3=Elisabetta |date=July 2019 |title=Resting state fMRI analysis using unsupervised learning algorithms | journal=
Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization |volume=8 |issue=3 | url=https://www.tandfonline.com/doi/abs/10.1080/21681163.2019.1636413 |publisher=Taylor&Francis |pages=2168–1171 |doi= 10.1080/21681163.2019.1636413}}{{Cite journal|last1=Burda|first1=Z|last2=Kornelsen|first2=J|last3=Nowak|first3=MA|last4=Porebski|first4=B|last5=Sboto-Frankenstein|first5=U|last6=Tomanek|first6=B|last7=Tyburczyk|first7=J|title=Collective Correlations of Brodmann Areas fMRI Study with RMT-Denoising|journal=Acta Physica Polonica B|date=2013|volume=44|issue=6|page=1243|doi=10.5506/APhysPolB.44.1243|arxiv=1306.3825|bibcode=2013AcPPB..44.1243B}}
).
=Optimal control=
In optimal control theory, the evolution of n state variables through time depends at any time on their own values and on the values of k control variables. With linear evolution, matrices of coefficients appear in the state equation (equation of evolution). In some problems the values of the parameters in these matrices are not known with certainty, in which case there are random matrices in the state equation and the problem is known as one of stochastic control.{{cite book |last=Chow |first=Gregory P. |year=1976 |title=Analysis and Control of Dynamic Economic Systems |location=New York |publisher=Wiley |isbn=0-471-15616-7 }}{{rp|ch. 13}}{{cite journal |last=Turnovsky |first=Stephen |year=1974 |title=The stability properties of optimal economic policies |journal=American Economic Review |volume=64 |issue=1 |pages=136–148 |jstor=1814888 }} A key result in the case of linear-quadratic control with stochastic matrices is that the certainty equivalence principle does not apply: while in the absence of multiplier uncertainty (that is, with only additive uncertainty) the optimal policy with a quadratic loss function coincides with what would be decided if the uncertainty were ignored, the optimal policy may differ if the state equation contains random coefficients.
=Computational mechanics=
In computational mechanics, epistemic uncertainties underlying the lack of knowledge about the physics of the modeled system give rise to mathematical operators associated with the computational model, which are deficient in a certain sense. Such operators lack certain properties linked to unmodeled physics. When such operators are discretized to perform computational simulations, their accuracy is limited by the missing physics. To compensate for this deficiency of the mathematical operator, it is not enough to make the model parameters random, it is necessary to consider a mathematical operator that is random and can thus generate families of computational models in the hope that one of these captures the missing physics. Random matrices have been used in this sense,{{Cite journal|last=Soize|first=C.|date=2005-04-08|title=Random matrix theory for modeling uncertainties in computational mechanics
|url=https://hal-upec-upem.archives-ouvertes.fr/hal-00686187/file/publi-2005-CMAME-194_12-16_1333-1366-soize-preprint.pdf|journal=Computer Methods in Applied Mechanics and Engineering|language=en|volume=194|issue=12–16|pages=1333–1366|doi=10.1016/j.cma.2004.06.038|bibcode=2005CMAME.194.1333S |s2cid=58929758 |issn=1879-2138}} with applications in vibroacoustics, wave propagations, materials science, fluid mechanics, heat transfer, etc.
=Engineering=
Random matrix theory can be applied to the electrical and communications engineering research efforts to study, model and develop Massive Multiple-Input Multiple-Output (MIMO) radio systems.{{Citation needed|date=May 2024}}
History
{{Expand section|date=April 2024}}
Random matrix theory first gained attention beyond mathematics literature in the context of nuclear physics. Experiments by Enrico Fermi and others demonstrated evidence that individual nucleons cannot be approximated to move independently, leading Niels Bohr to formulate the idea of a compound nucleus. Because there was no knowledge of direct nucleon-nucleon interactions, Eugene Wigner and Leonard Eisenbud approximated that the nuclear Hamiltonian could be modeled as a random matrix. For larger atoms, the distribution of the energy eigenvalues of the Hamiltonian could be computed in order to approximate scattering cross sections by invoking the Wishart distribution.{{Cite web |url=https://academic.oup.com/edited-volume/43656/chapter/365878918 |access-date=2024-04-22 |website=academic.oup.com |doi=10.1093/oxfordhb/9780198744191.013.2 |title=History – an overview |date=2015 |editor-last1=Akemann |editor-last2=Baik |editor-last3=Di Francesco |editor-first1=Gernot |editor-first2=Jinho |editor-first3=Philippe |last1=Bohigas |first1=Oriol |last2=Weidenmuller |first2=Hans |pages=15–40 |isbn=978-0-19-874419-1 }}
Gaussian ensembles
The most-commonly studied random matrix distributions are the Gaussian ensembles: GOE, GUE and GSE. They are often denoted by their Dyson index, β = 1 for GOE, β = 2 for GUE, and β = 4 for GSE. This index counts the number of real components per matrix element.
= Definitions =
The Gaussian unitary ensemble is described by the Gaussian measure with density
on the space of Hermitian matrices . Here
is a normalization constant, chosen so that the integral of the density is equal to one. The term unitary refers to the fact that the distribution is invariant under unitary conjugation. The Gaussian unitary ensemble models Hamiltonians lacking time-reversal symmetry.
The Gaussian orthogonal ensemble is described by the Gaussian measure with density
on the space of n × n real symmetric matrices H = (Hij){{su|b=i,j=1|p=n}}. Its distribution is invariant under orthogonal conjugation, and it models Hamiltonians with time-reversal symmetry. Equivalently, it is generated by , where is an matrix with IID samples from the standard normal distribution.
The Gaussian symplectic ensemble is described by the Gaussian measure with density
on the space of n × n Hermitian quaternionic matrices, e.g. symmetric square matrices composed of quaternions, {{math|1=H = (Hij){{su|b=i,j=1|p=n}}}}. Its distribution is invariant under conjugation by the symplectic group, and it models Hamiltonians with time-reversal symmetry but no rotational symmetry.
= Point correlation functions =
The ensembles as defined here have Gaussian distributed matrix elements with mean ⟨Hij⟩ = 0, and two-point correlations given by
from which all higher correlations follow by Isserlis' theorem.
= Moment generating functions =
The moment generating function for the GOE iswhere is the Frobenius norm.
= Spectral density =
File:Spectral density of gaussian ensembels, N = 1 to 32.png. The number of "humps" is equal to N.]]
The joint probability density for the eigenvalues {{math|λ1, λ2, ..., λn}} of GUE/GOE/GSE is given by
{{NumBlk||
where Zβ,n is a normalization constant which can be explicitly computed, see Selberg integral. In the case of GUE (β = 2), the formula (1) describes a determinantal point process. Eigenvalues repel as the joint probability density has a zero (of th order) for coinciding eigenvalues , and .
More succinctly, where is the Vandermonde determinant.
The distribution of the largest eigenvalue for GOE, and GUE, are explicitly solvable.{{cite journal |author=Chiani M| title=Distribution of the largest eigenvalue for real Wishart and Gaussian random matrices and a simple approximation for the Tracy-Widom distribution |journal=Journal of Multivariate Analysis |volume=129|pages=69–81|year=2014|doi=10.1016/j.jmva.2014.04.002|arxiv = 1209.3394 |s2cid=15889291}} They converge to the Tracy–Widom distribution after shifting and scaling appropriately.
= Convergence to Wigner semicircular distribution =
The spectrum, divided by , converges in distribution to the semicircular distribution on the interval : . Here is the variance of off-diagonal entries. The variance of the on-diagonal entries do not matter.
=Distribution of level spacings=
From the ordered sequence of eigenvalues , one defines the normalized spacings , where is the mean spacing. The probability distribution of spacings is approximately given by,
for the orthogonal ensemble GOE ,
for the unitary ensemble GUE , and
for the symplectic ensemble GSE .
The numerical constants are such that is normalized:
and the mean spacing is,
for .
Generalizations
Wigner matrices are random Hermitian matrices such that the entries
above the main diagonal are independent random variables with zero mean and have identical second moments.
The Gaussian ensembles can be extended for using the Dumitriu-Edelman tridiagonal ensemble.{{Citation |last1=Dumitriu |first1=Ioana |title=Matrix models for beta ensembles |date=2002-06-25 |id=arXiv:math-ph/0206043 |last2=Edelman |first2=Alan|journal=Journal of Mathematical Physics |volume=43 |issue=11 |pages=5830–5847 |doi=10.1063/1.1507823 |arxiv=math-ph/0206043 |bibcode=2002JMP....43.5830D }}
Invariant matrix ensembles are random Hermitian matrices with density on the space of real symmetric/Hermitian/quaternionic Hermitian matrices, which is of the form where the function {{math|V}} is called the potential.
The Gaussian ensembles are the only common special cases of these two classes of random matrices. This is a consequence of a theorem by Porter and Rosenzweig.{{Cite journal |last1=Porter |first1=C. E. |last2=Rosenzweig |first2=N. |date=1960-01-01 |title=STATISTICAL PROPERTIES OF ATOMIC AND NUCLEAR SPECTRA |url=https://www.osti.gov/biblio/4147616 |journal=Ann. Acad. Sci. Fennicae. Ser. A VI |language=English |volume=44|osti=4147616 }}{{Citation |last1=Livan |first1=Giacomo |title=Classified Material |date=2018 |url=https://doi.org/10.1007/978-3-319-70885-0_3 |work=Introduction to Random Matrices: Theory and Practice |pages=15–21 |editor-last=Livan |editor-first=Giacomo |access-date=2023-05-17 |series=SpringerBriefs in Mathematical Physics |place=Cham |publisher=Springer International Publishing |language=en |doi=10.1007/978-3-319-70885-0_3 |isbn=978-3-319-70885-0 |last2=Novaes |first2=Marcel |last3=Vivo |first3=Pierpaolo |volume=26 |editor2-last=Novaes |editor2-first=Marcel |editor3-last=Vivo |editor3-first=Pierpaolo}}
Spectral theory of random matrices
The spectral theory of random matrices studies the distribution of the eigenvalues as the size of the matrix goes to infinity. {{cite arXiv |last=Meckes |first=Elizabeth |title=The Eigenvalues of Random Matrices |date=2021-01-08 |class=math.PR |eprint=2101.02928}}
= Empirical spectral measure =
The empirical spectral measure {{math|μH}} of {{math|H}} is defined by
Usually, the limit of is a deterministic measure; this is a particular case of self-averaging. The cumulative distribution function of the limiting measure is called the integrated density of states and is denoted N(λ). If the integrated density of states is differentiable, its derivative is called the density of states and is denoted ρ(λ).
== Alternative expressions ==
= Types of convergence =
Given a matrix ensemble, we say that its spectral measures converge weakly to iff for any measurable set , the ensemble-average converges:Convergence weakly almost surely: If we sample independently from the ensemble, then with probability 1,for any measurable set .
In another sense, weak almost sure convergence means that we sample , not independently, but by "growing" (a stochastic process), then with probability 1,
for any measurable set .
For example, we can "grow" a sequence of matrices from the Gaussian ensemble as follows:
- Sample an infinite doubly infinite sequence of standard random variables
.
- Define each where is the matrix made of entries
.
Note that generic matrix ensembles do not allow us to grow, but most of the common ones, such as the three Gaussian ensembles, do allow us to grow.
=Global regime=
In the global regime, one is interested in the distribution of linear statistics of the form .
The limit of the empirical spectral measure for Wigner matrices was described by Eugene Wigner; see Wigner semicircle distribution and Wigner surmise. As far as sample covariance matrices are concerned, a theory was developed by Marčenko and Pastur..{{cite journal |last1=Marčenko |first1=V A |last2=Pastur |first2=L A |title=Distribution of eigenvalues for some sets of random matrices |journal=Mathematics of the USSR-Sbornik |volume=1 |issue=4 |year=1967 |doi=10.1070/SM1967v001n04ABEH001994 |pages=457–483 |bibcode = 1967SbMat...1..457M }}{{harvnb|Pastur|1973}}
The limit of the empirical spectral measure of invariant matrix ensembles is described by a certain integral equation which arises from potential theory.{{cite journal|last1=Pastur|first1=L.|last2=Shcherbina|first2 = M.|author2-link= Mariya Shcherbina |title=On the Statistical Mechanics Approach in the Random Matrix Theory: Integrated Density of States |journal=J. Stat. Phys. |year=1995 |volume=79|issue=3–4 |pages=585–611 |doi=10.1007/BF02184872|bibcode = 1995JSP....79..585D |s2cid=120731790}}
==Fluctuations==
For the linear statistics {{math|1=Nf,H = n−1 Σ f(λj)}}, one is also interested in the fluctuations about ∫ f(λ) dN(λ). For many classes of random matrices, a central limit theorem of the form
is known.{{cite journal| last=Johansson|first=K.| title=On fluctuations of eigenvalues of random Hermitian matrices|journal=Duke Math. J. | year=1998| volume=91| issue=1| pages=151–204 |doi=10.1215/S0012-7094-98-09108-6}}{{cite journal |last=Pastur|first=L.A. |title=A simple approach to the global regime of Gaussian ensembles of random matrices|journal=Ukrainian Math. J.| year=2005|volume=57|issue=6|pages=936–966 |doi=10.1007/s11253-005-0241-4 | s2cid=121531907| url=http://dspace.nbuv.gov.ua/handle/123456789/165749}}
== The variational problem for the unitary ensembles ==
Consider the measure
:
where is the potential of the ensemble and let be the empirical spectral measure.
We can rewrite with as
:
the probability measure is now of the form
:
where is the above functional inside the squared brackets.
Let now
:
be the space of one-dimensional probability measures and consider the minimizer
:
For there exists a unique equilibrium measure through the Euler-Lagrange variational conditions for some real constant
:
:
where is the support of the measure and define
:.
The equilibrium measure has the following Radon–Nikodym density
= Mesoscopic regime =
{{Cite journal |last1=Erdős |first1=László |last2=Schlein |first2=Benjamin |last3=Yau |first3=Horng-Tzer |date=April 2009 |title=Local Semicircle Law and Complete Delocalization for Wigner Random Matrices |url=http://link.springer.com/10.1007/s00220-008-0636-9 |journal=Communications in Mathematical Physics |language=en |volume=287 |issue=2 |pages=641–655 |arxiv=0803.0542 |bibcode=2009CMaPh.287..641E |doi=10.1007/s00220-008-0636-9 |issn=0010-3616}} The typical statement of the Wigner semicircular law is equivalent to the following statement: For each fixed interval
\Delta \lambda] centered at a point , as , the number of dimensions of the gaussian ensemble increases, the proportion of the eigenvalues falling within the interval converges to
\Delta \lambda]} \rho(t) dt, where is the density of the semicircular distribution.
If can be allowed to decrease as increases, then we obtain strictly stronger theorems, named "local laws" or "mesoscopic regime".
The mesoscopic regime is intermediate between the local and the global. In the mesoscopic regime, one is interested in the limit distribution of eigenvalues in a set that shrinks to zero, but slow enough, such that the number of eigenvalues inside
.
For example, the Ginibre ensemble has a mesoscopic law: For any sequence of shrinking disks with areas
inside the unite disk, if the disks have area
, the conditional distribution of the spectrum inside the disks also converges to a uniform distribution. That is, if we cut the shrinking disks along with the spectrum falling inside the disks, then scale the disks up to unit area, we would see the spectra converging to a flat distribution in the disks.{{Cite journal |last1=Bourgade |first1=Paul |last2=Yau |first2=Horng-Tzer |last3=Yin |first3=Jun |date=2014-08-01 |title=Local circular law for random matrices |url=https://doi.org/10.1007/s00440-013-0514-z |journal=Probability Theory and Related Fields |language=en |volume=159 |issue=3 |pages=545–595 |doi=10.1007/s00440-013-0514-z |issn=1432-2064|arxiv=1206.1449 }}
=Local regime=
In the local regime, one is interested in the limit distribution of eigenvalues in a set that shrinks so fast that the number of eigenvalues remains
.
Typically this means the study of spacings between eigenvalues, and, more generally, in the joint distribution of eigenvalues in an interval of length of order 1/n. One distinguishes between bulk statistics, pertaining to intervals inside the support of the limiting spectral measure, and edge statistics, pertaining to intervals near the boundary of the support.
==Bulk statistics==
Formally, fix in the interior of the support of . Then consider the point process
where are the eigenvalues of the random matrix.
The point process captures the statistical properties of eigenvalues in the vicinity of . For the Gaussian ensembles, the limit of is known; thus, for GUE it is a determinantal point process with the kernel
(the sine kernel).
The universality principle postulates that the limit of as should depend only on the symmetry class of the random matrix (and neither on the specific model of random matrices nor on ). Rigorous proofs of universality are known for invariant matrix ensembles{{cite journal
| last1 = Pastur | first1 = L.
| last2 = Shcherbina | first2 = M.|author2-link= Mariya Shcherbina
| title = Universality of the local eigenvalue statistics for a class of unitary invariant random matrix ensembles
| journal = Journal of Statistical Physics
| year = 1997
| volume = 86
| issue = 1–2
| pages = 109–147
| doi = 10.1007/BF02180200
| bibcode = 1997JSP....86..109P
| s2cid = 15117770
| url = http://purl.umn.edu/2773
| last1 = Deift | first1 = P.
| last2 = Kriecherbauer | first2 = T.
| last3 = McLaughlin | first3 = K.T.-R.
| last4 = Venakides | first4 = S.
| last5 = Zhou | first5 = X.
| title = Asymptotics for polynomials orthogonal with respect to varying exponential weights
| journal = International Mathematics Research Notices
| year = 1997
| issue = 16
| pages = 759–782
| doi = 10.1155/S1073792897000500 | volume=1997
| doi-access = free
}} and Wigner matrices.{{cite journal
| last1 = Erdős | first1 = L.
| last2 = Péché | first2 = S. | author2-link = Sandrine Péché
| last3 = Ramírez | first3 = J.A.
| last4 = Schlein | first4 = B.
| last5 = Yau | first5 = H.T.
| title = Bulk universality for Wigner matrices
| journal = Communications on Pure and Applied Mathematics
| year = 2010
| volume = 63
| issue = 7
| pages = 895–925
| doi = 10.1002/cpa.20317
| arxiv= 0905.4176
| last1 = Tao | first1 = Terence | authorlink1 = Terence Tao
| last2 = Vu | first2 = Van H. | authorlink2 = Van H. Vu
| title = Random matrices: universality of local eigenvalue statistics up to the edge
| journal = Communications in Mathematical Physics
| year = 2010
| volume = 298
| issue = 2
| pages = 549–572
| doi = 10.1007/s00220-010-1044-5
| bibcode = 2010CMaPh.298..549T
| arxiv = 0908.1982
| s2cid = 16594369 }}
==Edge statistics==
{{Main|Tracy–Widom distribution}}One example of edge statistics is the Tracy–Widom distribution.
As another example, consider the Ginibre ensemble. It can be real or complex. The real Ginibre ensemble has i.i.d. standard Gaussian entries , and the complex Ginibre ensemble has i.i.d. standard complex Gaussian entries .
Now let be sampled from the real or complex ensemble, and let be the absolute value of its maximal eigenvalue:We have the following theorem for the edge statistics:{{Cite journal |last=Rider |first=B |date=2003-03-28 |title=A limit theorem at the edge of a non-Hermitian random matrix ensemble |url=https://iopscience.iop.org/article/10.1088/0305-4470/36/12/331 |journal=Journal of Physics A: Mathematical and General |volume=36 |issue=12 |pages=3401–3409 |doi=10.1088/0305-4470/36/12/331 |bibcode=2003JPhA...36.3401R |issn=0305-4470}}
{{Math theorem
| name = Edge statistics of the Ginibre ensemble
| note =
| math_statement = For and as above, with probability one,
Moreover, if and
then converges in distribution to the Gumbel law, i.e., the probability measure on with cumulative distribution function .
}}
This theorem refines the circular law of the Ginibre ensemble. In words, the circular law says that the spectrum of almost surely falls uniformly on the unit disc. and the edge statistics theorem states that the radius of the almost-unit-disk is about , and fluctuates on a scale of , according to the Gumbel law.
Correlation functions
The joint probability density of the eigenvalues of random Hermitian matrices , with partition functions of the form
Z_n = \int_{M \in \mathbf{H}^{n \times n}} d\mu_0(M)e^{\text{tr}(V(M))}
where
V(x):=\sum_{j=1}^\infty v_j x^j
and is the standard Lebesgue measure on the space of Hermitian matrices, is given by
p_{n,V}(x_1,\dots, x_n) = \frac{1}{Z_{n,V}}\prod_{i
The -point correlation functions (or marginal distributions)
are defined as
R^{(k)}_{n,V}(x_1,\dots,x_k) = \frac{n!}{(n-k)!} \int_{\mathbf{R}}dx_{k+1} \cdots \int_{\R} dx_{n} \, p_{n,V}(x_1,x_2,\dots,x_n),
which are skew symmetric functions of their variables.
In particular, the one-point correlation function, or density of states, is
R^{(1)}_{n,V}(x_1) = n\int_{\R}dx_{2} \cdots \int_{\mathbf{R}} dx_{n} \, p_{n,V}(x_1,x_2,\dots,x_n).
Its integral over a Borel set gives the expected number of eigenvalues contained in :
\int_{B} R^{(1)}_{n,V}(x)dx = \mathbf{E}\left(\#\{\text{eigenvalues in }B\}\right).
The following result expresses these correlation functions as determinants of the matrices formed from evaluating the appropriate integral kernel at the pairs of points appearing within the correlator.
Theorem [Dyson-Mehta]
For any , the -point correlation function can be written as a determinant
R^{(k)}_{n,V}(x_1,x_2,\dots,x_k) = \det_{1\leq i,j \leq k}\left(K_{n,V}(x_i,x_j)\right),
where is the th Christoffel-Darboux kernel
K_{n,V}(x,y) := \sum_{k=0}^{n-1}\psi_k(x)\psi_k(y),
associated to , written in terms of the quasipolynomials
\psi_k(x) = {1\over \sqrt{h_k}}\, p_k(z)\, e^{- V(z) / 2} ,
where is a complete sequence of monic polynomials, of the degrees indicated, satisfying the orthogonilty conditions
\int_{\mathbf{R}} \psi_j(x) \psi_k(x) dx = \delta_{jk}.
Other classes of random matrices
=Wishart matrices=
{{Main|Wishart distribution}}
Wishart matrices are n × n random matrices of the form {{math|1=H = X X*}}, where X is an n × m random matrix (m ≥ n) with independent entries, and X* is its conjugate transpose. In the important special case considered by Wishart, the entries of X are identically distributed Gaussian random variables (either real or complex).
The limit of the empirical spectral measure of Wishart matrices was found by Vladimir Marchenko and Leonid Pastur.
=Random unitary matrices=
{{main|Circular ensembles}}
=Non-Hermitian random matrices=
{{main|Circular law}}
Selected bibliography
= Books =
- {{cite book|last=Mehta|first=M.L.| title=Random Matrices|year=2004| publisher=Elsevier/Academic Press |location=Amsterdam|isbn=0-12-088409-7}}
- {{cite book|last1=Anderson|first1=G.W.| last2=Guionnet|first2=A.|last3=Zeitouni|first3=O.|title=An introduction to random matrices.|year=2010|publisher=Cambridge University Press|location=Cambridge| isbn=978-0-521-19452-5}}
- {{cite book|last1=Akemann|first1=G.| last2=Baik|first2=J. |last3=Di Francesco|first3=P. | title=The Oxford Handbook of Random Matrix Theory |year=2011|publisher=Oxford University Press |location=Oxford| isbn=978-0-19-957400-1}}
- {{cite book |last1=Potters |first1=Marc |title=A First Course in Random Matrix Theory: for Physicists, Engineers and Data Scientists |last2=Bouchaud |first2=Jean-Philippe |date=2020-11-30 |publisher=Cambridge University Press |isbn=978-1-108-76890-0 |doi=10.1017/9781108768900}}
= Survey articles =
- {{cite journal|last1=Edelman|first1=A. |last2=Rao|first2=N.R|title=Random matrix theory| journal=Acta Numerica|year=2005| volume=14|pages=233–297 |doi=10.1017/S0962492904000236 |bibcode = 2005AcNum..14..233E |s2cid=16038147 }}
- {{cite journal|last=Pastur|first=L.A. | title=Spectra of random self-adjoint operators|journal=Russ. Math. Surv.| year=1973|volume=28|issue=1| pages=1–67 |doi=10.1070/RM1973v028n01ABEH001396 |bibcode = 1973RuMaS..28....1P |s2cid=250796916 }}
- {{cite journal | last1=Diaconis | first1=Persi | authorlink1=Persi Diaconis | title=Patterns in eigenvalues: the 70th Josiah Willard Gibbs lecture | mr=1962294 | year=2003 | journal=Bulletin of the American Mathematical Society |series=New Series | volume=40 | issue=2 | pages=155–178 | doi=10.1090/S0273-0979-03-00975-3| doi-access=free }}
- {{cite journal | last1=Diaconis | first1=Persi | authorlink1=Persi Diaconis | title=What is ... a random matrix? | url=https://www.ams.org/notices/200511/ | mr=2183871 | year=2005 | journal=Notices of the American Mathematical Society | issn=0002-9920 | volume=52 | issue=11 | pages=1348–1349}}
- {{Cite arXiv| last1=Eynard | first1=Bertrand | last2=Kimura | first2=Taro | last3=Ribault | first3=Sylvain | title=Random matrices | date=2015-10-15 | class=math-ph | eprint=1510.04430v2 }}
= Historic works =
- {{cite journal|last=Wigner|first=E.| title=Characteristic vectors of bordered matrices with infinite dimensions |journal=Annals of Mathematics |year=1955 |volume=62 |pages=548–564 | doi=10.2307/1970079| issue=3|jstor=1970079}}
- {{cite journal|last=Wishart|first=J.| title=Generalized product moment distribution in samples|journal=Biometrika| year=1928|volume=20A|issue=1–2|pages=32–52|doi=10.1093/biomet/20a.1-2.32}}
- {{cite journal|last1=von Neumann|first1=J. | last2=Goldstine|first2=H.H. |title=Numerical inverting of matrices of high order |journal=Bull. Amer. Math. Soc. |year=1947|volume=53|pages=1021–1099 | doi=10.1090/S0002-9904-1947-08909-6| issue=11|doi-access=free}}
References
{{Reflist|30em}}
External links
- {{cite journal |last=Fyodorov |first=Y. |title=Random matrix theory |journal=Scholarpedia |year=2011 |volume=6 |issue=3 |pages=9886 |doi=10.4249/scholarpedia.9886 |bibcode=2011SchpJ...6.9886F |doi-access=free }}
- {{cite web |last=Weisstein |first=E. W. |title=Random Matrix |publisher=Wolfram MathWorld |url=http://mathworld.wolfram.com/RandomMatrix.html}}
{{Matrix classes}}
{{Authority control}}
{{DEFAULTSORT:Random Matrix}}