Divergence Inequalities with Applications in Ergodic Theory

Akshay Bansal; Alice Zheng; Ian George

REVIEW 3 major objections 5 minor 90 references

Divergence Inequalities with Applications in Ergodic Theory

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read For ergodic Markov chains, most smooth f-divergences contract asymptotically at the efficiently computable χ² rate.

desk verdict Useful Taylor-based divergence inequalities and a clean upper bound on f-divergence contraction rates, but the advertised characterization of the asymptotic rate by the χ² coefficient overstates what the proof delivers for non-reversible chains. read the letter →

arxiv 2411.17241 v3 pith:FLNOCODN submitted 2024-11-26 cs.IT math.ITquant-ph

Ian George , Alice Zheng , Akshay Bansal This is my paper · ORCID

classification cs.ITmath.ITquant-ph MSC 94A1760J1062B10

keywords f-divergencesχ²divergencecontractioncoefficientsdataprocessinginequalityPinskerMarkovchainmixingtimesPetzergodictheory

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

Most smooth f-divergences can be sandwiched between constant multiples of the χ²-divergence once their second-derivative curvature is bounded. Iterating a contractive channel makes those curvature factors converge to f''(1), leaving the χ² contraction coefficient as the n-th-root rate for most f-divergences. Because the χ² coefficient is the squared maximal correlation of the channel, the rate is computable and is the fastest a divergence in this class can contract. This gives explicit mixing-time bounds and a quantum version for Petz f-divergences.

What carries the argument

The central object is the second-order Taylor integral representation of an f-divergence, $$D_f(p\|q)=\$int_0^{1}$(1-t)\sum_{i\in\operatorname{supp}(q)} $q_i^{{-1}}$ f''\left(1+t\left(\frac{p_i}{q_i}-1\right)\right)(p_i-q_i)^2\,dt,$$ which immediately gives $$\frac{\$kappa_f^{{\downarrow}}$(p,q)}{2}\$chi^{2}$(p\|q)\leq D_f(p\|q)\leq \frac{\$kappa_f^{{\uparrow}}$(p,q)}{2}\$chi^{2}$(p\|q),$$ where $\kappa_f^{\uparrow}$ and $\kappa_f^{\downarrow}$ are the maximum and minimum of f'' along the likelihood-ratio segments between p and q. All later results flow from this sandwich: contraction-coefficient ratios are bounded by χ² ratios times curvature ratios, and after n channel iterations the curvature ratios move to f''(1). The χ² side is then evaluated through the maximal-correlation identity $\eta_{\chi^2}(W,\pi)=\rho_m(X,Y)^2$, which is spectral and efficiently computable.

What would settle it

Numerically evaluate, for the 4-state noisy-typewriter chain of Example 48 or any small irreducible aperiodic chain, the quantities $\eta_f(W^n,\pi)^{1/n}$ for $f(t)=t\log t$ and $\eta_{\chi^2}(W,\pi)$ to high precision for large n, while tracking $\sup_p\kappa_f^{\uparrow}(W^n p,\pi)$ over starting distributions concentrated near the boundary of the simplex. If the supremum curvature does not converge to $f''(1)$, or if the n-th root stabilizes strictly above $\eta_{\chi^2}(W,\pi)$, the uniform-convergence premise is violated and Eq. (52) would need a correction term.

Watch

Extended reading notes

Core claim

The load-bearing result is that for a finite Markov chain W with unique stationary distribution π, and any twice continuously differentiable convex f with f(1)=0, f''(1)>0 and finite positive L_f, the asymptotic contraction exponent obeys $\lim_{n\to\infty}\eta_f(W^n,\pi)^{1/n}\leq \eta_{\chi^2}(W,\pi)$ whenever W is irreducible and aperiodic, or scrambling, or indecomposable with π of full support. Together with the known lower bound $\eta_f(W,\pi)\geq \eta_{\chi^2}(W,\pi)$, this identifies the input-dependent χ² contraction coefficient as the canonical asymptotic exponent for smooth f-divergences, and in reversible chains the bound is attained. The proof passes through new sandwiched inequalities $D_f(p\|q)\in[\kappa_f^{\downarrow}(p,q)/2,\kappa_f^{\uparrow}(p,q)/2]\chi^2(p\|q)$, then shows the curvature factors converge to f''(1) under channel iteration. The same mechanism yields computable mixing times in total variation and in f-divergence, and an analogue for Petz f-divergences on quantum mixing channels.

Load-bearing premise

The proof needs $\sup_{p\in\mathcal{P}(X)}\kappa_f^{\uparrow}(W^n p,W^n\pi)\to f''(1)$: the curvature factor must converge uniformly over all starting distributions so that its n-th root tends to one. The manuscript invokes the classical convergence theorem, which gives pointwise total-variation convergence, but does not supply a uniformity argument, and it does not show that indecomposability with full-support π implies aperiodicity. If uniformity fails, Eq. (52) acquires an extra subexponential or exponential factor and the stated rate is not the whole story.

Editorial extensions

If this is right

For irreducible aperiodic, scrambling, or indecomposable full-support stationary Markov chains, the asymptotic contraction rate of any smooth f-divergence is bounded by the χ² rate; for reversible chains the two rates coincide.
Because $\eta_{\chi^2}(W,\pi)$ is the squared maximal correlation, the universal rate is computable from the spectrum of the channel's centered joint matrix.
The refined convergence corollary gives explicit finite-n bounds on total-variation distance to stationarity in terms of $\eta_{\chi^2}(W,\pi)$, $\pi_{\min}$, and the target accuracy, without requiring irreducibility.
For f-divergences with concave $g(t)=(f(t)-f(0))/t$, mixing times obey the same $\log(1/\eta_{\chi^2})$ scaling up to constants.
The analogous theorem holds for Petz f-divergences on quantum mixing channels, with the Petz χ² contraction coefficient replacing the classical one, though no efficient computation is claimed.

Reading between the lines

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

A testable extension is the time-inhomogeneous case: iterating different channels should make the n-th-root rate the geometric mean of the per-step χ² contraction coefficients rather than a single spectral number.
A natural next step is to supply an explicit uniform-convergence argument for $\kappa_f^{\uparrow}(W^n p,W^n\pi)$ over the simplex; a Doeblin-coefficient bound would likely provide it.
On the quantum side, any future efficient estimate of the Petz χ² contraction coefficient would immediately turn Theorem 72 into an operational mixing-time tool, a computational question the paper leaves open.
The sandwich inequalities are not tied to Markov chains and could be used to estimate f-divergences under any iterative data-processing map, including approximate Bayesian updates and privacy mechanisms.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 useful core of this paper is the Taylor-based machinery in Sections III and IV. The Pinsker-style bounds in Tables I and II are clean, explicit, and genuinely convenient; the constants check out, and the method is more general than Gilardoni's in a real way. The derivation of divergence relations from an integral form of Taylor's theorem is straightforward, reproducible, and clearly explained. I'd happily point a student at Theorem 31 and Corollary 33 as a tidy way to bound f-divergences by χ². The quantum lift to Petz f-divergences is mostly a corollary ride on the Nussbaum-Skoła representation, but it is honest about what it does and does not give: they never claim efficiency there, and they flag it.

The soft spot is the central ergodic claim. Theorem 50 proves an upper bound: limsup of the nth root of η_f(W^n,π) is at most η_{χ²}(W,π). The lower bound imported from [26] is one-step, and applying it to W^n and taking nth roots only gives a liminf bound in terms of the asymptotic spectral radius of the normalized operator. For non-reversible chains, that spectral radius can be strictly smaller than η_{χ²}(W,π), as the three-state scrambling example in the stress-test shows. So the abstract's phrase "rate of contraction ... is characterized by the input-dependent contraction coefficient of the χ²-divergence" is too strong. The theorem supports a one-sided rate statement, not a characterization. That is a real gap between the advertised result and the proved result, and it is not just a missing uniformity argument. The uniformity issue the reader flagged is secondary; under the stated hypotheses that part can be patched, whereas the missing lower bound is a false conclusion for non-reversible chains.

The paper also leans on the authors' prior ITW work without, as far as I can tell, clearly flagging the degree of overlap in the divergence-inequality core. That matters for a journal submission, though it is fixable by disclosure rather than by changing the math.

Who is this for? People working on contraction coefficients, strong data processing inequalities, and computable mixing-time bounds. The divergence inequalities are a solid reference contribution. The ergodic characterization needs to be restated as an upper bound or restricted to the reversible case where the tightness actually holds. With that revision, this deserves a serious referee.

Referee Report

3 major / 5 minor

Summary. The paper develops Taylor-expansion-based inequalities for twice-differentiable f-divergences: Pinsker-type lower bounds (Theorems 7 and 20) and two-sided bounds in terms of the chi-squared divergence (Theorem 31). These tools are then applied to input-dependent contraction coefficients of time-homogeneous Markov chains. The central advertised result, Theorem 50, asserts that for many f-divergences the asymptotic contraction rate is characterized by the input-dependent chi-squared contraction coefficient, which is efficiently computable. The paper also extends the divergence inequalities and ergodic applications to Petz f-divergences in quantum information theory, though without an efficient computation guarantee.

Significance. If the main application were correct as stated, the paper would be a substantial unification: most smooth f-divergence contraction rates would be governed by the efficiently computable chi-squared coefficient, with clean consequences for mixing times. The underlying Pinsker and chi-squared comparison inequalities in Sections III and IV are clearly useful, appear to be proved correctly, and generalize several known results. The quantum extension in Section VI is also of interest, even though computability is explicitly not guaranteed. However, the central characterization claim in the abstract and Section V is not supported by Theorem 50, which proves only an upper bound; for non-reversible chains the actual asymptotic rate can be strictly smaller than the chi-squared contraction coefficient. This overclaim is load-bearing and must be corrected or qualified before the paper can be accepted.

major comments (3)

[Abstract; Section V, Theorem 50, Eq. (44)] The central claim that eta_{chi^2}(W,pi) characterizes the asymptotic contraction rate is not established and is false as stated. Theorem 50 establishes at most limsup_n eta_f(W^n,pi)^{1/n} <= eta_{chi^2}(W,pi); for non-reversible chains the true rate can be strictly smaller. For the scrambling chain W = [[0.9,0.1,0],[0,0.8,0.2],[0.4,0,0.6]] with stationary distribution pi=(4,2,1)/7, the normalized operator A = D^{1/2} W D^{-1/2} has ||A||^2 = eta_{chi^2}(W,pi) ≈ 0.654, while the squared modulus of the non-unit eigenvalues is 0.44, so lim_n eta_{chi^2}(W^n,pi)^{1/n} = 0.44 for the chi^2-divergence itself. Thus the rate is not characterized by eta_{chi^2}(W,pi) except in the reversible case. The abstract's 'characterized by' and Section I's 'scale at a rate given by' must be replaced by an upper-bound statement, with tightness explicitly restricted to reversible chains.
[Section V, proof of Theorem 50, condition 3] The proof that condition 3 (W indecomposable, pi full support) implies convergence of W^n p to pi is invalid. The two-state swap chain W = [[0,1],[1,0]] with pi=(1/2,1/2) is indecomposable with full-support stationary distribution, but eta_{chi^2}(W,pi)=1 and W^n p does not converge for p ≠ pi. Inequalities (47)-(49) only yield convergence when eta_{chi^2}(W,pi)<1, which is not implied by the stated hypotheses. Consequently, the assertion 'Thus, in all cases, lim_n W^n p = pi' is false. Moreover, the passage to lim_n sup_{p} kappa^up_f(W^n p, W^n pi) = f''(1) in Eqs. (50)-(52) requires uniform convergence over p in P(X); for conditions 1 and 2 this can be recovered from uniform TV bounds, but for condition 3 it is not established. The theorem should either add eta_{chi^2}(W,pi)<1 (or an equivalent aperiodicity-type assumption) and prove the uniformity, or remove condition 3.
[Section V, Theorem 50 statement] The theorem states a limit, lim_{n to infinity} eta_f(W^n,pi)^{1/n} <= eta_{chi^2}(W,pi), but the proof establishes at most a limsup inequality: the sequence eta_f(W^n,pi)^{1/n} need not converge for every f and W. The statement and the limit laws used in Eqs. (50)-(52) should be rewritten with limsup; otherwise the theorem asserts existence of a limit that is not proved.

minor comments (5)

[Section V, proof of Lemma 49] In the displayed chain of inequalities, the symbol 'L_f ,1' appears in the denominator; it should be 'L_f'.
[Section V, Example 48] The text says the chain 'increases the value by one modulo 3' for a four-state alphabet; this should be 'modulo 4'.
[Section VI, Theorem 72] The sentence 'Moreover, we know the above bound can be tight given Theorem 50 and Proposition 65' is not justified as stated: Proposition 65 only reduces the quantum contraction coefficient to the classical one for classical-to-classical channels, and tightness in Theorem 50 is proved only for reversible Markov chains. The claim should be qualified or removed.
[Section III, Theorem 7] The constant L_f appears in the assumptions before it is defined; the statement should explicitly say that L_f is a positive real number satisfying condition (8) or (9).
[Section V, proof of Theorem 50, Eq. (50)] In Eq. (50) the exponent on eta_f(W^n,pi) is displayed as n rather than 1/n; the displayed limit should read lim_n eta_f(W^n,pi)^{1/n}.

Circularity Check

0 steps flagged · score 2.0 of 10

No structurally circular step: Taylor-derived inequalities support the bounds; minor self-citations are not load-bearing.

full rationale

The central derivation chain is self-contained. Section IV obtains the f-divergence/χ² inequalities from multivariate Taylor's theorem (Lemma 28, Lemma 30, Theorem 31), so the constants κ↑, κ↓ and the χ² factors are not fitted to the target contraction result. Lemma 49 and Theorem 50 then combine those inequalities with submultiplicativity and the externally cited lower bound η_f(W,π) ≥ η_χ²(W,π) from [26]; the n-th root limit is taken from the resulting expressions rather than assumed. The only self-citations are [77], which the paper actually disputes in a remark, and [71], an attribution of a Bregman Pinsker derivation; neither carries the proof of Theorem 50. The abstract's word 'characterized' is stronger than the theorem's 'at most' plus reversible-tightness, and the skeptic's non-reversible example is a correctness/over-claim issue, not a circular reduction. The proof gap noted by the reader (uniformity in sup κ↑→f''(1)) is likewise a missing justification, not a circular input. Score 2 reflects only the presence of minor non-load-bearing self-citations.

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

The paper's central claims depend on standard Taylor calculus, the data processing inequality, and prior contraction-coefficient results. It introduces no free parameters, no fitted constants, and no new physical or mathematical entities.

assumptions (6)

standard math Taylor's theorem with integral remainder (univariate and multivariate)
Used to derive all integral representations in Sections III and IV (Lemmas 5, 18, 28).
standard math Data processing inequality for classical f-divergences and operator-convex Petz f-divergences
Fact 4.1 and Fact 60; ensures monotonicity and enables lifting Pinsker inequalities to quantum (Corollary 62).
domain assumption Known properties of contraction coefficients: η_f ≥ η_{χ^2} for f''(1)>0, maximal correlation is efficiently computable, submultiplicativity
Proposition 41 items 3-4 and Proposition 64, cited from [26,27,74,75], used to sandwich η_f and to claim computability.
domain assumption Operator convexity of f is needed for Petz f-divergences to satisfy DPI
Fact 60; used in quantum extensions and in Proposition 69.
domain assumption Finite state space / finite Hilbert space assumptions
Section II states almost entire work focuses on finite dimensions; used for efficient computation and existence of stationary distributions.
domain assumption In Theorem 50, conditions (irreducible+aperiodic, scrambling, or indecomposable with full-support π) ensure convergence and η_{χ^2}<1
These are sufficient conditions for the rate result, but the paper does not prove the indecomposable case implies aperiodicity.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Divergence Inequalities with Applications in Ergodic Theory." pith.science (2026). https://pith.science/paper/FLNOCODN

@misc{pith2026241117241,
  author       = {Pith},
  title        = {Pith review of: Divergence Inequalities with Applications in Ergodic Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FLNOCODN}},
  note         = {Machine review of arXiv:2411.17241}
}

abstract

The data processing inequality is central to information theory and motivates the study of monotonic divergences. However, it is not clear operationally we need to consider all such divergences. We establish a simple method for Pinsker inequalities as well as general bounds in terms of $\chi^{2}$-divergences for twice-differentiable $f$-divergences. These tools imply new relations for input-dependent contraction coefficients. We use these relations to show for many $f$-divergences the rate of contraction of a time homogeneous Markov chain is characterized by the input-dependent contraction coefficient of the $\chi^{2}$-divergence. This is efficient to compute and the fastest it could converge for a class of divergences. We show similar ideas hold for mixing times. Moreover, we extend these results to the Petz $f$-divergences in quantum information theory, albeit without any guarantee of efficient computation. These tools may have applications in other settings where iterative data processing is relevant.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

90 extracted references · 72 canonical work pages

[26]

Strong data-processing inequalities for channels and bayesian networks

Yury Polyanskiy and Yihong Wu. Strong data-processing inequalities for channels and bayesian networks. In Convexity and Concentration , pages 211–249. Springer, 2017

work page 2017
[1]

A mathematical theory of communication

Claude Shannon. A mathematical theory of communication . Bell System T echnical Journal , 27(3): 379–423, 1948

1948
[2]

Data processing theorems and the second law of thermodynamics

Neri Merhav . Data processing theorems and the second law of thermodynamics. IEEE T ransactions on Information Theory , 57(8):4926–4939, 2011

2011
[3]

On information-type measure of differen ce of probability distributions and indirect observations

Imre Csisz´ ar. On information-type measure of differen ce of probability distributions and indirect observations. Studia Sci. Math. Hungar. , 2:299–318, 1967

1967
[4]

Expectations and entropy inequalitie s for ﬁnite quantum systems

G ¨ oran Lindblad. Expectations and entropy inequalitie s for ﬁnite quantum systems. Communications in Mathematical Physics , 39:111–119, 1974

1974
[5]

Quasi-entropies for ﬁnite quantum systems

D´ enes Petz. Quasi-entropies for ﬁnite quantum systems . Reports on Mathematical Physics , 23(1):57–65, 1986

1986
[6]

Sandwiched R´ enyi divergence satisﬁes da ta processing inequality

Salman Beigi. Sandwiched R´ enyi divergence satisﬁes da ta processing inequality . Journal of Mathematical Physics , 54(12), 2013

2013
[7]

Quantum Information Processing with Finite Resources

Marco T omamichel. Quantum Information Processing with Finite Resources . Springer International Publishing, 2016. doi: 10.1007/978-3-319-21891-5. 34

doi:10.1007/978-3-319-21891-5 2016

Show all 90 references

[8]

Entropy and relative en tropy from information-theoretic principles

Gilad Gour and Marco T omamichel. Entropy and relative en tropy from information-theoretic principles. IEEE T ransactions on Information Theory , 67(10):6313–6327, 2021

2021
[9]

Cover and Joy A

Thomas M. Cover and Joy A. Thomas. Elements of Information Theory . John Wiley & Sons, Inc., 2006. ISBN 0471241954

2006
[10]

Quantum information theory

Mark M Wilde. Quantum information theory . Cambridge University Press, 2013. doi: 10.1017/ CBO9781139525343

2013
[11]

Principles of quantum co mmunication theory: A modern approach

Sumeet Khatri and Mark M Wilde. Principles of quantum co mmunication theory: A modern approach. arXiv preprint arXiv:2011.04672 , 2020

2011 arXiv
[12]

Information Theory: From Coding to Learning

Yury Polyanskiy and Yihong Wu. Information Theory: From Coding to Learning . Cambridge University Press, 2025

2025
[13]

f -divergence inequalities

Igal Sason and Sergio V erd ´ u. f -divergence inequalities. IEEE T ransactions on Information Theory , 62 (11):5973–6006, 2016

2016
[14]

Approximating csiszar f-divergence by the use of taylor’s formula with integral re mainder

Neil S Barnett, Pietro Cerone, Sever Silvestru Dragomi r, and Anthony Sofo. Approximating csiszar f-divergence by the use of taylor’s formula with integral re mainder. Mathematical Inequalities and Applications, 5:417–434, 2002

2002
[15]

A reﬁnement and a divided difference r everse of Jensen’s inequality with applications

S Sever Dragomir. A reﬁnement and a divided difference r everse of Jensen’s inequality with applications. Revista Colombiana de Matem´ aticas, 50(1):17–39, 2016

2016
[16]

On f-divergences: Integral representatio ns, local behavior, and inequalities

Igal Sason. On f-divergences: Integral representatio ns, local behavior, and inequalities. Entropy, 20 (5):383, 2018

2018
[17]

Approximation of f-divergence measu res by using two points taylor’s type representations with integral remainders

Sever S Dragomir. Approximation of f-divergence measu res by using two points taylor’s type representations with integral remainders. Bulletin of the T ransilvania University of Brasov. Series I II: Mathematics, Informatics, Physics , 12(1):21–40, 2019

2019
[18]

M. S. Pinsker. Information and information stability o f random variables and processes (in russian). Izv. Akad. Nauk , 1960

1960
[19]

Note on discrimination information and var iation (corresp.)

Igor V ajda. Note on discrimination information and var iation (corresp.). IEEE T ransactions on Information Theory , 16(6):771–773, 1970

1970
[20]

Generalised pinske r inequalities

Mark D Reid and Robert C Williamson. Generalised pinske r inequalities. arXiv preprint arXiv:0906.1244, 2009

2009 arXiv
[21]

On Pinsker’s and Vajda’s type ineq ualities for csisz´ ar’s f -divergences

Gustavo L Gilardoni. On Pinsker’s and Vajda’s type ineq ualities for csisz´ ar’s f -divergences. IEEE T ransactions on Information Theory , 56(11):5377–5386, 2010

2010
[22]

Spreading of sets in pr oduct spaces and hypercontraction of the Markov operator

Rudolf Ahlswede and Peter G´ acs. Spreading of sets in pr oduct spaces and hypercontraction of the Markov operator. The annals of probability , pages 925–939, 1976

1976
[23]

Rautu, Mary Beth Ruskai, Eug ene Seneta, and Gh

Joel E Cohen, Y oh Iwasa, Gh. Rautu, Mary Beth Ruskai, Eug ene Seneta, and Gh. Zbaganu. Relative entropy under mappings by stochastic matrices. Linear algebra and its applications , 179:211–235, 1993

1993
[24]

Equ ivalence of certain entropy contraction coefﬁcients

Man-Duen Choi, Mary Beth Ruskai, and Eugene Seneta. Equ ivalence of certain entropy contraction coefﬁcients. Linear algebra and its applications , 208:29–36, 1994

1994
[25]

Strong data processing inequalities a nd phi-sobolev inequalities for discrete channels

Maxim Raginsky . Strong data processing inequalities a nd phi-sobolev inequalities for discrete channels. IEEE T ransactions on Information Theory , 62(6):3355–3389, 2016

2016
[27]

Comparison of contract ion coefﬁcients for f-divergences

Anuran Makur and Lizhong Zheng. Comparison of contract ion coefﬁcients for f-divergences. Problems of Information T ransmission , 56:103–156, 2020

2020
[28]

Strong data process ing constant is achieved by binary inputs

Or Ordentlich and Yury Polyanskiy . Strong data process ing constant is achieved by binary inputs. IEEE T ransactions on Information Theory , 68(3):1480–1481, 2021

2021
[29]

On contraction coefﬁcients, partial orders and approximation of capacities for quantum channel s

Christoph Hirche, Cambyse Rouz´ e, and Daniel Stilck Fr anc ¸a. On contraction coefﬁcients, partial orders and approximation of capacities for quantum channel s. Quantum, 6:862, 2022

2022
[30]

A lower bound on the space overhead of fault-tolerant quantum computation

Omar Fawzi, Alexander M ¨ uller-Hermes, and Ala Shayegh i. A lower bound on the space overhead of fault-tolerant quantum computation. arXiv preprint arXiv:2202.00119 , 2022

2022 arXiv
[31]

Quantum R´ enyi and f-divergences from integral representations, 2024

Christoph Hirche and Marco T omamichel. Quantum R´ enyi and f-divergences from integral representations, 2024

2024
[32]

Deﬁnition of random variab les by conditional distributions

Roland L’vovich Dobrushin. Deﬁnition of random variab les by conditional distributions. T eoriya V eroyatnostei i ee Primeneniya , 15(3):469–497, 1970

1970
[33]

Broadcasting on trees and the ising model

William Evans, Claire Kenyon, Yuval Peres, and Leonard J Schulman. Broadcasting on trees and the ising model. Annals of Applied Probability , pages 410–433, 2000. 35

2000
[34]

Local privacy and statistical minimax rates

John C Duchi, Michael I Jordan, and Martin J W ainwright. Local privacy and statistical minimax rates. In 2013 IEEE 54th annual symposium on foundations of computer s cience, pages 429–438. IEEE, 2013

2013
[35]

Dissipation of informat ion in channels with input constraints

Yury Polyanskiy and Yihong Wu. Dissipation of informat ion in channels with input constraints. IEEE T ransactions on Information Theory , 62(1):35–55, 2015

2015
[36]

Quantum differential privacy: An information theory perspective

Christoph Hirche, Cambyse Rouz´ e, and Daniel Stilck Fr anc ¸a. Quantum differential privacy: An information theory perspective. IEEE T ransactions on Information Theory , 69(9):5771–5787, 2023

2023
[37]

Quant um pufferﬁsh privacy: A ﬂexible privacy framework for quantum systems

Theshani Nuradha, Ziv Goldfeld, and Mark M Wilde. Quant um pufferﬁsh privacy: A ﬂexible privacy framework for quantum systems. IEEE T ransactions on Information Theory , 2024

2024
[38]

Quasi-entropies for states of a von Neuman n algebra

D´ enes Petz. Quasi-entropies for states of a von Neuman n algebra. Publications of the Research Institute for Mathematical Sciences , 21(4):787–800, 1985

1985
[39]

A lower bound for discrimination inf ormation in terms of variation (corresp.)

Solomon Kullback. A lower bound for discrimination inf ormation in terms of variation (corresp.). IEEE transactions on Information Theory , 13(1):126–127, 1967

1967
[40]

On the optimum rate of transmitt ing information

Johannes HB Kemperman. On the optimum rate of transmitt ing information. The Annals of Mathematical Statistics , pages 2156–2177, 1969

1969
[41]

Reﬁnements of Pinsker’s inequality

Alexei A Fedotov , Peter Harremo¨ es, and Flemming T opso e. Reﬁnements of Pinsker’s inequality . IEEE T ransactions on Information Theory , 49(6):1491–1498, 2003

2003
[42]

A distributi on dependent reﬁnement of Pinsker’s inequality

Erik Ordentlich and Marcelo J W einberger. A distributi on dependent reﬁnement of Pinsker’s inequality . IEEE T ransactions on Information Theory , 51(5):1836–1840, 2005

2005
[43]

On the minimum f-divergence for gi ven total variation

Gustavo L Gilardoni. On the minimum f-divergence for gi ven total variation. Comptes rendus. Math´ ematique, 343(11-12):763–766, 2006

2006
[44]

on the min imum f-divergence for given total variation

Gustavo L Gilardoni. Corrigendum to the note “on the min imum f-divergence for given total variation”[cr acad. sci. paris, ser. i 343 (2006) 763–766]. Comptes Rendus. Math´ ematique, 348(5-6): 299, 2010

2006
[45]

On divergences and info rmations in statistics and information theory

Friedrich Liese and Igor V ajda. On divergences and info rmations in statistics and information theory . IEEE T ransactions on Information Theory , 52(10):4394–4412, 2006

2006
[46]

On applications of the repres entation of f-divergences as averaged minimal Bayesian risk

Cornelius Gutenbrunner. On applications of the repres entation of f-divergences as averaged minimal Bayesian risk. In T ransactions of the 11th Prague Conference on Information T heory, Statistical Decision Functions and Random Processes , pages 449–456, 1992

1992
[47]

Statistical in formation and discrimination

Ferdinand Osterreicher and Igor V ajda. Statistical in formation and discrimination. IEEE transactions on information theory , 39(3):1036–1039, 1993

1993
[48]

Information theor y and statistics: A tutorial

Imre Csisz´ ar, Paul C Shields, et al. Information theor y and statistics: A tutorial. Foundations and T rends® in Communications and Information Theory , 1(4):417–528, 2004

2004
[49]

On the chi square and hig her-order chi distances for approximating f-divergences

Frank Nielsen and Richard Nock. On the chi square and hig her-order chi distances for approximating f-divergences. IEEE Signal Processing Letters , 21(1):10–13, 2013

2013
[50]

On asymptotic properties of information-theoretic divergences

Mdel C Pardo and Igor V ajda. On asymptotic properties of information-theoretic divergences. IEEE T ransactions on Information Theory , 49(7):1860–1867, 2003

2003
[51]

The χ 2-divergence and mixing times of quantum Markov processes

Kristan T emme, Michael James Kastoryano, Mary Beth Rus kai, Michael Marc W olf, and Frank V erstraete. The χ 2-divergence and mixing times of quantum Markov processes. Journal of Mathematical Physics , 51(12), 2010

2010
[52]

On relations betwee n the relative entropy and χ 2-divergence, generalizations and applications

T omohiro Nishiyama and Igal Sason. On relations betwee n the relative entropy and χ 2-divergence, generalizations and applications. Entropy, 22(5):563, 2020

2020
[53]

Context tree estimatio n for not necessarily ﬁnite memory processes, via bic and mdl

Imre Csisz´ ar and Zsolt T alata. Context tree estimatio n for not necessarily ﬁnite memory processes, via bic and mdl. IEEE T ransactions on Information theory , 52(3):1007–1016, 2006

2006
[54]

A note on reverse Pinsker inequalitie s

Olivier Binette. A note on reverse Pinsker inequalitie s. IEEE T ransactions on Information Theory , 65 (7):4094–4096, 2019

2019
[55]

Theshani Nuradha and Mark M. Wilde. Contraction of priv ate quantum channels and private quantum hypothesis testing, 2024. URL https://arxiv .org/ abs/2406.18651

2024
[56]

Sample complexity of locally differen- tially private quantum hypothesis testing, 2024

Hao-Chung Cheng, Christoph Hirche, and Cambyse Rouz´ e . Sample complexity of locally differen- tially private quantum hypothesis testing, 2024. URL https ://arxiv .org/abs/2406.18658

2024 arXiv
[57]

Geometric bounds fo r eigenvalues of Markov chains

Persi Diaconis and Daniel Stroock. Geometric bounds fo r eigenvalues of Markov chains. The annals of applied probability , pages 36–61, 1991

1991
[58]

Eigenvalue bounds on convergence to s tationarity for nonreversible Markov chains, with an application to the exclusion process

James Allen Fill. Eigenvalue bounds on convergence to s tationarity for nonreversible Markov chains, with an application to the exclusion process. The annals of applied probability , pages 62–87, 1991. 36

1991
[59]

Contraction of genera lized relative entropy under stochastic mappings on matrices

D´ enes Petz and Mary Beth Ruskai. Contraction of genera lized relative entropy under stochastic mappings on matrices. Inﬁnite Dimensional Analysis, Quantum Probability and Rel ated T opics, 1(01): 83–89, 1998

1998
[60]

Optimized quantum f-divergences and data processing

Mark M Wilde. Optimized quantum f-divergences and data processing. Journal of Physics A: Mathematical and Theoretical , 51(37):374002, 2018

2018
[61]

Beyond strong subadditivity? improv ed bounds on the contraction of generalized relative entropy

Mary Beth Ruskai. Beyond strong subadditivity? improv ed bounds on the contraction of generalized relative entropy . Reviews in Mathematical Physics , 6(05a):1147–1161, 1994

1994
[62]

Monotone riema nnian metrics and relative entropy on noncommutative probability spaces

Andrew Lesniewski and Mary Beth Ruskai. Monotone riema nnian metrics and relative entropy on noncommutative probability spaces. Journal of Mathematical Physics , 40(11):5702–5724, 1999

1999
[63]

Rever se-type data processing inequality , 2024

Paula Belzig, Li Gao, Graeme Smith, and Peixue Wu. Rever se-type data processing inequality , 2024. URL https://arxiv .org/abs/2411.19890

2024
[64]

Sa ndwiched R´ enyi convergence for quantum evolutions

Alexander M ¨ uller-Hermes and Daniel Stilck Franca. Sa ndwiched R´ enyi convergence for quantum evolutions. Quantum, 2:55, 2018

2018
[65]

One-variable calculus, with an introduction to linear alge bra, volume 1

T om M Apostol. One-variable calculus, with an introduction to linear alge bra, volume 1. Blaisdell W altham, T oronto, London, 1967

1967
[66]

V ector calculus

Jerrold E Marsden and Anthony Tromba. V ector calculus. Macmillan, 2003

2003
[67]

The Analysis of Linear Partial Differential Operators I

Lars H ¨ ormander. The Analysis of Linear Partial Differential Operators I . Springer-V erlag, 2nd edition, 2003

2003
[68]

Information contraction and decomposition

Anuran Makur. Information contraction and decomposition . PhD thesis, Massachusetts Institute of T echnology , 2019

2019
[69]

Relative informat ion of type s, csisz´ ar’s f-divergence, and information inequalities

Inder Jeet T aneja and Pranesh Kumar. Relative informat ion of type s, csisz´ ar’s f-divergence, and information inequalities. Information sciences , 166(1-4):105–125, 2004

2004
[70]

Bounds on f-divergences and related distan ces

Igal Sason. Bounds on f-divergences and related distan ces. CCIT Report 859, T echnion — Israel Institute of T echnology , 2014

2014
[71]

Online learning of a panoply of quantum objects (in submission), 2024

Akshay Bansal, Ian George, Soumik Ghosh, Jamie Sikora, and Alice Zheng. Online learning of a panoply of quantum objects (in submission), 2024

2024
[72]

W eak ergodicity in n on-homogeneous Markov chains

John Hajnal and Maurice S Bartlett. W eak ergodicity in n on-homogeneous Markov chains. Mathematical Proceedings of the Cambridge Philosophical S ociety, 54(2):233–246, 1958

1958
[73]

Non-negative matrices and Markov chains

Eugene Seneta. Non-negative matrices and Markov chains . Springer Science & Business Media, 2006

2006
[74]

On sequences of pairs of dependent random variables

Hans S Witsenhausen. On sequences of pairs of dependent random variables. SIAM Journal on Applied Mathematics , 28(1):100–113, 1975

1975
[75]

A new data processing inequa lity and its applications in distributed source and channel coding

W ei Kang and Sennur Ulukus. A new data processing inequa lity and its applications in distributed source and channel coding. IEEE T ransactions on Information Theory , 57(1):56–69, 2010

2010
[76]

Markov chains and mixing times , volume 107

David A Levin and Yuval Peres. Markov chains and mixing times , volume 107. American Mathematical Soc., 2017

2017
[77]

Divergence inequalities from multivariate taylor’s theorem

Ian George, Alice Zheng, and Akshay Bansal. Divergence inequalities from multivariate taylor’s theorem. In Information Theory Workshop 2024 , 2024

2024
[78]

Approximations of the s pectral radius corresponding eigenvector, and second largest modulus of an eigenvalue for square, nonn egative, irreducible matrices

Orna Gross and Uriel G Rothblum. Approximations of the s pectral radius corresponding eigenvector, and second largest modulus of an eigenvalue for square, nonn egative, irreducible matrices. SIAM journal on matrix analysis and applications , 14(1):15–32, 1993

1993
[79]

The Theory of Quantum Information

John W atrous. The Theory of Quantum Information . Cambridge University Press, 2018

2018
[80]

Quantum f-divergences and error correction

Fumio Hiai, Mil´ an Mosonyi, D´ enes Petz, and C´ edric B´ eny . Quantum f-divergences and error correction. Reviews in Mathematical Physics , 23(07):691–747, 2011

2011
[81]

Contraction coefﬁcien ts for noisy quantum channels

Fumio Hiai and Mary Beth Ruskai. Contraction coefﬁcien ts for noisy quantum channels. Journal of Mathematical Physics , 57(1), 2016

2016
[82]

Different quantum f-div ergences and the reversibility of quantum operations

Fumio Hiai and Mil´ an Mosonyi. Different quantum f-div ergences and the reversibility of quantum operations. Reviews in Mathematical Physics , 29(07):1750023, 2017

2017
[83]

A new quantum version of f-divergence

Keiji Matsumoto. A new quantum version of f-divergence . In Masanao Ozawa, Jeremy Butterﬁeld, Hans Halvorson, Mikl ´ os R´ edei, Yuichiro Kitajima, and Fra ncesco Buscemi, editors, Reality and Measurement in Algebraic Quantum Theory , pages 229–273, Singapore, 2018. Springer S...

2018
[84]

The Chernoff lower bound for symmetric quantum hypothesis testing

Michael Nussbaum and Arleta Szkoła. The Chernoff lower bound for symmetric quantum hypothesis testing. The Annals of Statistics , 37(2):1040–1057, 2009. ISSN 00905364, 21688966. URL http://www .jstor.org/stable/30243657. 37

2009
[85]

A f ully quantum asymptotic equipartition property

Marco T omamichel, Roger Colbeck, and Renato Renner. A f ully quantum asymptotic equipartition property . IEEE T ransactions on information theory , 55(12):5840–5847, 2009

2009
[86]

Matrix analysis , volume 169

Rajendra Bhatia. Matrix analysis , volume 169. Springer Science & Business Media, 1997

1997
[87]

Quantum channels and operations-guide d tour, 2012

Michael M W olf. Quantum channels and operations-guide d tour, 2012. URL https://mediatum.ub.tum.de/doc/1701036/1701036.pdf

2012
[88]

Ergodic and mixing quantum channels in ﬁnite dimensions

Daniel Burgarth, Giulio Chiribella, Vittorio Giovann etti, Paolo Perinotti, and Kazuya Yuasa. Ergodic and mixing quantum channels in ﬁnite dimensions. New Journal of Physics , 15(7):073045, 2013

2013
[89]

T ensorization of the strong data processing inequality for quantum chi- square divergences

Yu Cao and Jianfeng Lu. T ensorization of the strong data processing inequality for quantum chi- square divergences. Quantum, 3:199, 2019

2019
[90]

Quantum Doeblin coefﬁcients: A simp le upper bound on contraction coefﬁcients

Christoph Hirche. Quantum Doeblin coefﬁcients: A simp le upper bound on contraction coefﬁcients. In 2024 IEEE International Symposium on Information Theory (I SIT), pages 557–562, 2024. doi: 10.1109/ ISIT57864.2024.10619667. APPENDIX A LOWER BOUNDS FOR f -D IVERGENCES Proof of...

2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [26]

Strong data-processing inequalities for channels and bayesian networks

Yury Polyanskiy and Yihong Wu. Strong data-processing inequalities for channels and bayesian networks. In Convexity and Concentration , pages 211–249. Springer, 2017

work page 2017

[2] [1]

A mathematical theory of communication

Claude Shannon. A mathematical theory of communication . Bell System T echnical Journal , 27(3): 379–423, 1948

1948

[3] [2]

Data processing theorems and the second law of thermodynamics

Neri Merhav . Data processing theorems and the second law of thermodynamics. IEEE T ransactions on Information Theory , 57(8):4926–4939, 2011

2011

[4] [3]

On information-type measure of differen ce of probability distributions and indirect observations

Imre Csisz´ ar. On information-type measure of differen ce of probability distributions and indirect observations. Studia Sci. Math. Hungar. , 2:299–318, 1967

1967

[5] [4]

Expectations and entropy inequalitie s for ﬁnite quantum systems

G ¨ oran Lindblad. Expectations and entropy inequalitie s for ﬁnite quantum systems. Communications in Mathematical Physics , 39:111–119, 1974

1974

[6] [5]

Quasi-entropies for ﬁnite quantum systems

D´ enes Petz. Quasi-entropies for ﬁnite quantum systems . Reports on Mathematical Physics , 23(1):57–65, 1986

1986

[7] [6]

Sandwiched R´ enyi divergence satisﬁes da ta processing inequality

Salman Beigi. Sandwiched R´ enyi divergence satisﬁes da ta processing inequality . Journal of Mathematical Physics , 54(12), 2013

2013

[8] [7]

Quantum Information Processing with Finite Resources

Marco T omamichel. Quantum Information Processing with Finite Resources . Springer International Publishing, 2016. doi: 10.1007/978-3-319-21891-5. 34

doi:10.1007/978-3-319-21891-5 2016