Open Access This article is
- freely available
Entropy 2018, 20(9), 641; https://doi.org/10.3390/e20090641
Rényi Entropy Power Inequalities via Normal Transport and Rotation
LTCI, Télécom ParisTech, Université Paris-Saclay, 75013 Paris, France
École Polytechnique, Université Paris-Saclay, 91128 Palaiseau, France
Received: 7 July 2018 / Accepted: 23 August 2018 / Published: 26 August 2018
Following a recent proof of Shannon’s entropy power inequality (EPI), a comprehensive framework for deriving various EPIs for the Rényi entropy is presented that uses transport arguments from normal densities and a change of variable by rotation. Simple arguments are given to recover the previously known Rényi EPIs and derive new ones, by unifying a multiplicative form with constant c and a modification with exponent of previous works. In particular, for log-concave densities, we obtain a simple transportation proof of a sharp varentropy bound.
Keywords:Rényi entropy; entropy power inequalities; transportation arguments; normal distributions; escort distributions; log-concave distributions
The entropy power inequality (EPI) dates back to Shannon’s seminal paper  and has a long history . The link with the Rényi entropy was first made by Dembo, Cover and Thomas  in connection with Young’s convolutional inequality with sharp constants, where Shannon’s EPI is obtained by letting the Rényi entropy orders tend to one ( Theorem 17.8.3).
The Rényi entropy  was first defined as a generalization of Shannon’s entropy for discrete variables, when looking for the most general definition of information measures that would preserve the additivity for independent events. It has found many applications such as source coding , hypothesis testing , channel coding  and guessing . The (differential) Rényi entropy considered in this paper (Definition 2) generalizes the (differential) Shannon’s entropy for continuous variables. It was first considered in  to make the transition between the entropy-power and the Brunn–Minkowski inequalities. It has also been applied to deconvolution problems . A definition of the Renyi entropy-power itself appears in , which is essentially Definition 5 below.
Recently, there has been significant interest in Rényi entropy power inequalities for several independent variables (the survey  is recommended to the reader for recent developments on forward and reverse EPIs). Bobkov and Chistyakov  extended the classical Shannon’s EPI to the Rényi entropy by incorporating a multiplicative constant that depends on the order of the Rényi entropy. Ram and Sason  improved the value of the constant by making it depend also on the number of variables. Bobkov and Marsiglietti  proved another modification of the EPI for the Rényi entropy for two independent variables, with a power exponent parameter whose value was further improved by Li . All these EPIs were found for Rényi entropies of orders >1. The -modification of the Rényi EPI was extended to orders <1 for two independent variables having log-concave densities by Marsiglietti and Melbourne . The starting point of all the above works was Young’s strengthened convolutional inequality.
Recently, Shannon’s original EPI was given a simple proof  using a transport argument from normal variables and a change of variable by rotation. In this paper, we exploit these ingredients, described in the following lemmas, to establish all the above-mentioned Rényi EPIs and derive new ones.
Throughout this article, the considered n-dimensional zero-mean random variables admit a density which is implicitly assumed continuous inside its support. We write if X has density f and write if is normally distributed with covariance matrix .
Lemma 1 (Normal Transport).
Let . There exists a diffeomorphism such that . Moreover, T can be chosen such that its Jacobian matrix is (lower) triangular with positive diagonal elements.
Lemma 1 is known in optimal transport theory as an application of the Knothe–Rosenblatt map [19,20]. Two different proofs are given in . The proof is very simple for one-dimensional variables , where T is just an increasing function with continuous derivative .
(Normal Rotation ). If are i.i.d. (independent and identically distributed) and normal, then for any , the rotationyields i.i.d. normal variables .
Notice that the starred variables can be expressed in terms of the tilde variables by the inverse rotation
The proof of Lemma 2 is trivial considering covariance matrices. A deeper result states that this property of remaining i.i.d. by rotation characterizes the normal distribution—this is known as Bernstein’s lemma (see, e.g., ( [Lemma 4]) and ( [Chapter 5])). This explains why one obtains equality in the EPI only for normal variables (see  for more details).
This article is a revised, full version of what was presented in part in a previous conference communication . It is organized as follows. Preliminary definitions and known properties are presented in Section 2. Section 3 derives a crucial “information inequality” for Rényi entropies that enjoys a transformational invariance. The central result is in Section 4, where the first version of the Rényi EPI by Dembo, Cover and Thomas is proven using the ingredients of Lemmas 1 and 2. All previously known Rényi EPIs for finite orders—and new ones—are then derived using a simple method in Section 5. Section 6 concludes.
2. Preliminary Definitions and Properties
Throughout this article, we consider exponents with . The following definition is well known and used, e.g., in Hölder’s inequality.
Definition 1 (Conjugate Exponent).
The conjugate exponent of p isthat is, the number such that
There are two situations depending on whether is positive or negative, as summarized in the following table:
Definition 2 (Rényi Entropy).
If X has density , its Rényi entropy of order p is defined bywhere denotes the norm of f.
It is known that the limit as is the Shannon entropy
The Rényi entropy enjoys well-known properties similar to those of the Shannon entropy, which are recalled here for completeness.
Lemma 3 (Scaling Property).
For any ,
Making a change of variables, . ☐
One recovers the usual scaling property for the Shannon entropy by letting .
Lemma 4 (Rényi Entropy of the Normal).
If for some nonsingular covariance matrix , thenwhere denotes the determinant. In particular, for ,
By direct calculation, . ☐
Again, one recovers the Shannon entropy of a normal variable by letting (then, ).
(Escort Density ( § 2.2)). If , its escort density of exponent p is the density defined byIn other words, , where denotes the norm of f. We also use the notation to denote the corresponding escort random variable with density .
Lemma 5 (Monotonicity Property).
If , then with equality if and only if X is uniformly distributed.
Let and assume that for all q in a neighborhood of p so that one can freely differentiate under the integral sign:where denotes the Kullback–Leibler divergence. Equality can hold only if a.e., which, since , implies that f is constant over some measurable subset , and is zero elsewhere. It follows that for any if X is not uniformly distributed. Conversely, if X is uniformly distributed over some measurable subset , its density can be written as for all and elsewhere. Then, is independent of p. ☐
Notice the identity established in the proof:
A similar formula for discrete variables can be found in ( § 5.3).
3. An Information Inequality
The Shannon entropy satisfies a fundamental “information inequality” ( Theorem 2.6.3) from which many classical information-theoretic inequalities can be derived. This can be written asfor any density , with equality if and only if a.e. The following Theorem 1 can be seen as the natural extension of the information inequality to Rényi entropies and is central in the following derivations of this paper. Bercher  pointed out to the author that it is similar to an inequality for discrete distributions established by Campbell  in the context of source coding (see also ).
Theorem 1 (Information Inequality).
For any density φ,with equality if and only if a.e.
By letting , one recovers the classical information inequality in Equation (13) for the Shannon entropy.
By definition in Equation (5),where the inequality follows from Jensen’s inequality applied to the function , which is strictly concave if (that is, ) and strictly convex if (that is, ). Equality holds if and only if is constant a.e., which means that and are proportional a.e. Normalizing gives the announced condition a.e. ☐
An alternate proof is obtained using Hölder’s inequality or its reverse applied to f and . Notice that the equality case for givesas can be easily checked directly.
The following conditional version of Theorem 1 involves a more complicated relation for dependent variables.
Corollary 1 (Conditional Information Inequality)
For any two random variables ,where denotes the Rényi entropy of X knowing and the expectation on the l.h.s. is taken over Y (the expectation in the r.h.s. is taken over ).
In particular, when X and Y are independent,with equality if and only if does not depend on y and equals a.e.
From Equation (14) for fixed y, one has for () and the opposite inequality for (), with equality if and only if a.e. Taking the expectation over Y yields for () and the opposite inequality for (). The result follows by taking the logarithm and multiplying by . When X and Y are independent, equality holds if and only if a.e. for all y. ☐
For the Shannon entropy, the difference between the two sides of the information inequality in Equation (13) is the Kullback–Leibler divergence:which can also be noted where and . It is known (and easy to check) that the divergence is invariant by reversible transformations T. This means that, when and , one has . A natural extension to Rényi entropies can be obtained on the differencebetween the two sides of the information inequality in Equation (14).
Theorem 2 (Transformational Invariance).
Let be a diffeomorphism and suppose thatwhere and . Then,
Note that, from Equation (5), this identity can be rewritten as
Proceed to prove Equation (23). Let be the respective densities of and recall that and . By the transformation T, the densities are related bywhere denotes the Jacobian determinant of T. Using these relations and Definition 3,
φ is a density and is not used in the proof of Theorem 2. Therefore, Equation (22) holds more generally for any function φ satisfying Equation (25).
4. First Version of the Rényi EPI
For two independent random variables X and Y, the Shannon entropy power inequality can be expressed as follows [2,3]: For any ,with equality if and only if are i.i.d. normal—more precisely (given the translation invariance of the entropy) when they are normal with identical covariance matrices; since it was assumed in Notation 1 that all considered variables have zero mean, both statements are equivalent. This means that the difference is minimum (zero) for i.i.d. normal . In this section, we study the natural generalization for Rényi entropies ( Theorem 12), namely that the quantityis minimum for i.i.d. normal. Here, the triple and its associated satisfy the following condition, which is used, e.g., in Young’s convolutional inequality.
Definition 4 (Exponent Triple).
An exponent triple has conjugates of the same sign and such that
The corresponding coefficient is defined by
In other words, the exponents are such thatand fulfill one the following two conditions:
The key argument used in this section is the following. If and , then for the escort variables, and . By normal transport (Lemma 1), one can writefor two diffeomorphims T and U, where are, say, i.i.d. standard normal . (It follows that and .) We then have the following straightforward extension of Theorem 2.
Lemma 6 (Transformational Invariance for Two Independent Variables).
For a two-dimensional ,where
By the transformations T and U, the densities of the escort variables are related by and . Now, by the same calculation as in the proof of Theorem 2,
Let φ be the density of . Then,
By the equality case of Theorem 1 (see (16)), one has
Now, by Lemma 6 applied to , we have , and, therefore,
Since from Lemma 1, T and U can be chosen such that and are (lower) triangular with positive diagonal elements, it follows easily from the arithmetic-geometric mean inequality that
The result follows at once (for either positive or negative ). ☐
We can now use the normal rotation Lemma 2 to conclude by proving the following,
(Rényi EPI ). For independent and exponent triple ,with equality if and only if are i.i.d. normal.
If are i.i.d. normal, then is also identically distributed as X and Y, and from Lemma 4, it is immediate to check that equality holds (irrespective of their covariances). Therefore, the inequality in Equation (42) is equivalent towhere , are, say, i.i.d. standard normal .
To prove Equation (43), consider the normal rotation of Lemma 2 and write in terms of using Equation (2) in the first term of the r.h.s. of Equation (38) (Lemma 7). One obtains:where
Making the change of variable , one obtainssince is a density. Hence, is also a density in for fixed . Now, since by Lemma 2, and are independent, by the conditional information inequality in Equation (18) of Corollary 1, one has
It remains to settle the equality case in Equation (43). From the above proof, equality holds in Equation (43) if and only if both Equation (41) and Equation (47) are equalities. Equality in Equation (41) holds if and only if for all ,
Since and are independent normal variables, this implies that and are constant and equal. In particular the Jacobian in (45) is constant.
From Corollary 1 equality in Equation (47) holds if and only if does not depend on , which implies that does not depend on the value of . Taking derivatives with respect to for all ,which impliesfor all . Therefore, T and U are linear transformations, equal up to an additive constant (equal to 0 since all variables are assumed of zero mean). It follows that and are normal with respective distributions and . Hence, X and Y are i.i.d. normal . ☐
A straightforward generalization to several independent variables is the following.
Corollary 2 (Rényi EPI for Several Variables).
Let be exponents; those conjugates are of the same sign and satisfyand let be defined by
Then, for independent ,with equality if and only if the are i.i.d. normal.
By induction on m: The result for is Theorem 3. Suppose the result satisfied at order and let and be such that . Notice that , hence . By Theorem 3, with equality if and only if are i.i.d. normal. Now, by the induction hypothesis, with equality if and only if the ()—and hence —are i.i.d. normal. The result at order m follows by combining the two inequalities since . ☐
5. Recent Versions of the Rényi EPI
(Rényi Entropy Power ). The Rényi entropy power of order r is defined by
Up to a multiplicative constant, is the (average) power of a white normal variable having the same Rényi entropy as X—hence the name “entropy power”. In fact, if has the same Rényi entropy , then by Lemma 4,
The Renyi entropy power enjoys the same scaling property as for the usual power: By Lemma 3, for any ,
For independent , Rényi entropy power inequalities take either the form [13,14]for some positive constant c, or the form [15,16,17]for some positive exponent . The constants c and may depend on the order r, the number m of variables and the dimension n. What is desired is:
- a maximum possible value of c in Equation (57) since the inequality is automatically satisfied for all positive constants .
- a minimum possible value of in Equation (58) since the inequality is automatically satisfied for all positive exponents ; in fact, since Equation (58) is homogeneous by scaling the variables as in Equation (56), one may suppose without loss of generality that the r.h.s. of Equation (58) is ; then, , hence for all i and .
The following useful characterization, which generalizes ( Lemma 2.1), makes the link between the various versions (Equations (53), (57), and (58)) of the Rényi entropy power inequality.
For independent , the Rényi EPI in the general formfor some constant and exponent is equivalent to the following inequalityfor any positive such that , where denotes the discrete entropy
5.1. Rényi Entropy Power Inequalities for Orders >1
From Lemma 8 and Corollary 2, it is easy to recover known Rényi EPIs and obtain new ones for orders . In fact, if , then and all are positive and . Therefore, all are and by monotonicity (Lemma 5),
Plugging this into Equation (53), one obtainswhere for . For future reference, define(The absolute value is needed in the next subsection where is negative.)
This function is strictly convex in because is strictly convex. Note that vanishes in the limiting cases where tends to one of the standard unit vectors , , …, and since every is a convex combination of these vectors and is strictly convex, one has .
By Lemma 8 for we only need to check that the r.h.s. of Equation (74) is greater than for any choice of the ’s, that is, for any choice of exponents such that . Thus, Equation (57) will hold for . Now, by the log-sum inequality ( Theorem 2.7.1),with equality if and only if all are equal, that is, the are equal to . Thus, which yields . ☐
An alternate proof is to argue that is convex and symmetrical in and is, therefore, minimized when all are equal.
The above constant c is certainly not optimal since equality in Equation (73) holds if and only if the are uniformly distributed (Lemma 5) while equality in Equation (53) holds if and only if the are identically normally distributed (Corollary 2). Ram and Sason  tightened Equation (57) further using optimization techniques, resulting in a constant that depends on the relative values of the entropy powers themselves.
It can be noted that decreases (and tends to ) as m increases; in fact . Thus, a universal constant independent of m is obtained by taking
This was the constant established by Bobkov and Chistyakov .
The Rényi EPI (58) holds for and and this value of α cannot be improved using the method of this paper by making it depend on m.
By Lemma 8, for , we only need to check that the r.h.s. of Equation (74) is greater than for any choice of s, that is, for any choice of exponents such that . Thus, Equation (58) will hold for . By the proof of the preceding theorem, the numerator is minimized when all are equal and this also maximizes the entropy in the denominator. However, one cannot conclude yet since the minimum in the numerator is negative.
A stationary point is easily obtained by the Lagrangian method, which implies that is constant independent of i. This gives that is constant, hence a stationary point is obtained when all are equal (to ) and the corresponding value of is .
However, the boundary of the domain of is the simplex where on each vertex joining two standard unit vectors, has the same expression as for . Li  showed—this is also easily proved using [17, Lemma 8]—that, for , the minimum is obtained when . The corresponding value of is , which is easily seen to be less than for any .
Therefore, the minimum of is attained at the boundary when all are zero except two of them equal to . This gives . ☐
The case yields which was found by Li  who remarked that this value of α is strictly smaller (better) than the value obtained by Bobkov and Marsiglietti .
Interestingly, for , the exponent of Theorem 5 cannot be improved by this method. In fact, in the above proof, it is easily seen that is negative and increases toward 0 as m increases. Therefore, the exponent α cannot be decreased (improved) as m increases.
The above value of is . However, using the same method, it is easy to obtain Rényi EPIs with exponent values . This is given by the following Theorem 6.
The Rényi EPI (59) holds for , with .
By Lemma 8 we only need to check that the r.h.s. of Equation (74) is greater than , that is, for any choice of s, that is, for any choice of exponents such that . Thus, for a given , Equation (59) will hold for . From the preceding proofs (since both and are convex functions of ), the minimum is attained when all s are equal. This gives . ☐
5.2. Rényi Entropy Power Inequalities for Orders <1 and Log-Concave Densities
If , then and all are negative and . Therefore, all are and by monotonicity (Lemma 5), the opposite inequality of Equation (73) holds and the method of the preceding subsection fails. For log-concave densities, however, Equation (73) can be replaced by a similar inequality in the right direction.
Definition 6 (Log-Concave Density).
A density f is log-concave if is concave in it support, i.e., for all ,
If X has a log-concave density, then is concave in p.
As noted below in Corollary 3, this is essentially a result obtained by Fradelizi, Madiman and Wang . The following alternate proof uses the transport properties seen in Section 4.
Define where . By Lemma 1 there exists two diffeomorphisms such that one can write and . Then, has density
Now, by log-concavity, Equation (79) with ,
Using the arithmetic-geometric mean inequality in Equation (41) and integrating the density in Equation (80) over , one obtains
Taking the logarithm yields the announced concavity. ☐
As a side result, it is interesting to note that we have obtained a simple transportation proof of the following varentropy bound:
(Varentropy Bound ). One has , that is, .
Since is concave, one has , that is,
Differentiating twice using Leibniz’s product rule and plugging the identity in Equation (12), the l.h.s. of this inequality becomes
As another easy consequence of Lemma 9, since is concave and vanishes for , the slopes are non-increasing in p. In other words, is nondecreasing. Therefore:
(Marsiglietti and Melbourne ). If then for any X with log-concave density, .
We can now use Lemma 8 and Corollary 2 to obtain Rényi EPIs for orders . Since all are , by Corollary 4,
Plugging this into Equation (53), one obtainswhere we have used that for . Notice that the r.h.s. of Equation (89) for () is the opposite of that of Equation (74) for (). However, since is now negative, the r.h.s. is exactly equal to which is still convex and negative. For this reason, the proofs of the following theorems for are such repeats of the theorems obtained previously for .
The Rényi EPI (57) for log-concave densities holds for and .
It is identical to that of Theorem 4 except for the change in the expression of . ☐
The Rényi EPI (58) for log-concave densities holds for and and this value of α cannot be improved using the method of this paper by making it depend on m.
It is identical to that of Theorem 5 except for the change in the expression of . ☐
The case yields which was found by Marsiglietti and Melbourne . Again, the exponent of the theorem is not improved for .
The Rényi EPI in Equation (59) for log-concave densities holds for where and .
It is identical to that of Theorem 6 except for the change in the expression of . ☐
This article provides a comprehensive framework to derive known Rényi entropy power inequalities (with shorter proofs), and prove new ones. The framework is based on a transport argument from normal densities and a change of variable by rotation. Only basic properties of Rényi entropies are used in the proofs.
In particular, the -modification of the EPI is recovered for two or more independent variables for Rényi entropy orders >1 as well as for orders <1, and the Rényi EPI with multiplicative constant c is extended to Rényi entropy orders <1. In addition, a more general formulation with both exponent and constant c is obtained for all orders. In passing, a simple proof using normal transport of a recent sharp varentropy bound was obtained for log-concave densities.
As a perspective, the methods developed in this paper can perhaps be generalized to obtain reverse Rényi entropy power inequalities (see, e.g., the discussion in ).
This research received no external funding.
Conflicts of Interest
The author declares no conflict of interest.
- Shannon, C.E. A Mathematical Theory of Communication. Bell Syst. Tech. J. 1948, 27, 623–656. [Google Scholar] [CrossRef]
- Rioul, O. Information Theoretic Proofs of Entropy Power Inequalities. IEEE Trans. Inf. Theory 2011, 57, 33–55. [Google Scholar] [CrossRef][Green Version]
- Dembo, A.; Cover, T.M.; Thomas, J.A. Information Theoretic Inequalities. IEEE Trans. Inf. Theory 1991, 37, 1501–1518. [Google Scholar] [CrossRef]
- Cover, T.M.; Thomas, J.A. Elements of Information Theory, 2nd ed.; Wiley: Hoboken, NJ, USA, 2006. [Google Scholar]
- Rényi, A. On Measures of Information and Entropy. In Proceedings of the Fourth Berkeley Symposium on Mathematics, Statistics and Probability, Berkeley, CA, USA, 20 June–30 July 1960; University of California Press: Berkeley, CA, USA, 1960; Volume 1, pp. 547–561. [Google Scholar]
- Campbell, L.L. A Coding Theorem and Rényi’s Entropy. Inf. Control 1965, 8, 423–429. [Google Scholar] [CrossRef]
- Ben-Bassat, M.; Raviv, J. Rényi’s Entropy and the Probability of Error. IEEE Trans. Inf. Theory 1978, 24, 324–331. [Google Scholar] [CrossRef]
- Arimoto, S. Information Measures and Capacity of Order α for Discrete Memoryless Channels. In Topics in Information Theory, Colloquia mathematica Societatis János Bolyai 16; Csiszár, I., Elias, P., Eds.; North Holland: Amsterdam, The Netherlands, 1977; pp. 41–52. [Google Scholar]
- Arikan, E. An Inequality on Guessing and its Application to Sequential Decoding. IEEE Trans. Inf. Theory 1996, 42, 99–105. [Google Scholar] [CrossRef][Green Version]
- Erdogmus, D.; Hild, K.E.; Principe, J.C.; Lazaro, M.; Santamaria, I. Adaptive Blind Deconvolution of Linear Channels Using Rényi’s Entropy with Parzen Window Estimation. IEEE Trans. Signal Process. 2004, 52, 1489–1498. [Google Scholar] [CrossRef]
- Savaré, G.; Toscani, G. The Concavity of Rényi Entropy Power. IEEE Trans. Inf. Theory 2014, 60, 2687–2693. [Google Scholar] [CrossRef]
- Madiman, M.; Melbourne, J.; Xu, P. Forward and Reverse Entropy Power Inequalities in Convex Geometry. In Convexity and Concentration; Carlen, E., Madiman, M., Werner, E., Eds.; IMA Volumes in Mathematics and its Applications; Springer: Berlin, Germany, 2017; Volume 161, pp. 427–485. [Google Scholar]
- Bobkov, S.G.; Chistyakov, G.P. Entropy Power Inequality for the Rényi Entropy. IEEE Trans. Inf. Theory 2015, 61, 708–714. [Google Scholar] [CrossRef]
- Ram, E.; Sason, I. On Rényi Entropy Power Inequalities. IEEE Trans. Inf. Theory 2016, 62, 6800–6815. [Google Scholar] [CrossRef][Green Version]
- Bobkov, S.G.; Marsiglietti, A. Variants of the Entropy Power Inequality. IEEE Trans. Inf. Theory 2017, 63, 7747–7752. [Google Scholar] [CrossRef][Green Version]
- Li, J. Rényi Entropy Power Inequality and a Reverse. Stud. Math. 2018, 242, 303–319. [Google Scholar] [CrossRef]
- Marsiglietti, A.; Melbourne, J. On the Entropy Power Inequality for the Rényi Entropy of Order [0, 1]. arxiv, 2018; arXiv:1710.00800. [Google Scholar]
- Rioul, O. Yet Another Proof of the Entropy Power Inequality. IEEE Trans. Inf. Theory 2017, 63, 3595–3599. [Google Scholar] [CrossRef][Green Version]
- Rosenblatt, M. Remarks on a Multivariate Transformation. Ann. Math. Stat. 1952, 23, 470–472. [Google Scholar] [CrossRef]
- Knothe, H. Contributions to the theory of convex bodies. Mich. Math. J. 1957, 4, 39–52. [Google Scholar] [CrossRef]
- Rioul, O. Optimal Transportation to the Entropy-Power Inequality. In Proceedings of the IEEE Information Theory and Applications Workshop (ITA 2017), San Diego, CA, USA, 12–17 February 2017. [Google Scholar]
- Bryc, W. The Normal Distribution—Characterizations with Applications; Lecture Notes in Statistics; Springer: Berlin, Germany, 1995. [Google Scholar]
- Rioul, O. Optimal Transport to Rényi Entropies. In Proceedings of the 3rd Conference on Geometric Science of Information (GSI 2017), Paris, France, 7–9 November 2017; Lecture Notes in Computer Science. Springer: Berlin, Germany, 2017. [Google Scholar]
- Bercher, J.F. Source Coding with Escort Distributions and Rényi Entropy Bounds. Phys. Lett. A 2009, 373, 3235–3238. [Google Scholar] [CrossRef][Green Version]
- Bercher, J.F. On Generalized Cramér-Rao Inequalities, Generalized Fisher Information and Characterizations of Generalized q-Gaussian Distributions. J. Phys. A Math. Theor. 2012, 45, 255303. [Google Scholar] [CrossRef][Green Version]
- Beck, C.; Schlögl, F. Thermodynamics of Chaotic Systems: An Introduction; Cambridge University Press: Cambridge, UK, 1993. [Google Scholar]
- Bercher, J.F.; (ESIEE Paris, Marne-la-Vallée, France). Private communication, 2018.
- Fradelizi, M.; Madiman, M.; Wang, L. Optimal Concentration of Information Content for Log-Concave Densities. In High Dimensional Probability VII: The Cargèse Volume; Houdré, C., Mason, D.M., Reynaud-Bouret, P., Rosiński, J., Eds.; Birkhäuser: Basel, Switzerland, 2016. [Google Scholar]
© 2018 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).