On The Quantum Renyi Relative Entropies and Their Use

On the quantum Rényi relative entropies and their use
Mark M. Wilde
Hearne Institute for Theoretical Physics,

Department of Physics and Astronomy,
Center for Computation and Technology,
Louisiana State University,
Baton Rouge, Louisiana, USA
[email protected]
On Sabbatical with
Stanford Institute for Theoretical Physics,
Stanford University,
Stanford, California 94305
ISL Colloquium, Stanford University, Stanford, California, September 17, 2020
Mark M. Wilde (LSU) 1 / 68

The phrase “and their use” and its use
I noticed that many people in quantum information theory like to use the
phrase “and their use” or the words “and its use” specifically at the end of
their book or paper titles, so I decided to be part of this trend:
Denes Petz, “Quantum entropy and its use” (1991)
Denes Petz, “From f -divergence to quantum quasi-entropies and their

use” (2009)
Felix Leditzky, “Relative entropies and their use in quantum

information theory” (2016)
M. E. Shirokov, “Upper bounds for the Holevo quantity and their

use” (2017)
I also added “On the” at the front to make the title even more appealing
to academics...

Motivation
Entropy and relative entropy are at the core of information theory
Rényi entropy and Rényi relative entropy are fruitful

generalizations of this concept, having many applications
When generalizing to the quantum case, there is not a unique way to

generalize Rényi relative entropy
In this talk, I’ll review three quantum generalizations of Rényi relative

entropy and their information-processing properties
I’ll also discuss their applications in quantum information theory

Entropy
One of the most famous formulas in all of science [Sha48]:
X
H(pX ) := − pX (x) log2 pX (x)
x
Figure: Screen capture from the movie Arrival (2016)
Equal to the optimal limit of classical data compression [Sha48]

Relative entropy
Relative entropy [KL51] generalizes entropy
Let pX ∈ P(X ) be a probability distribution, and let qX be a

measure. Then

X pX (x)
D(pX kqX ) := pX (x) log2
x
qX (x)
Distinguishability measure with interpretation in hypothesis testing

Special cases of relative entropy
Entropy is a special case of relative entropy:
H(pX ) = −D(pX k1),
where 1 is the vector of all ones.
Mutual information is a special case of relative entropy:
I (X ; Y ) := D(pXY kpX ⊗ pY ) = inf D(pXY kpX ⊗ qY )

qY ∈P(Y)

Data-processing inequality
Let NY |X be a classical channel and define
qY := NY |X (qX )
where X
qY (y ) = NY |X (y |x)qX (x).
x∈X
Then the data-processing inequality for relative entropy is
D(pX kqX ) ≥ D(NY |X (pX )kNY |X (qX ))

Rényi relative entropy
Rényi relative entropy is a generalization of relative entropy with

parameter α ∈ (0, 1) ∪ (1, ∞) [Rén61]:
!
1 X
α 1−α
Dα (pX kqX ) := log2 pX (x) qX (x)
α−1
x∈X

Special cases of Rényi relative entropy
Rényi entropy [Rén61] is a special case:

!
1 X
Hα (pX ) := log2 pX (x)α = −Dα (pX k1)
1−α
x∈X
Rényi mutual information [Csi95] defined as
Iα (X ; Y ) := inf Dα (pXY kpX ⊗ qY )

qY ∈P(Y)

Properties of Rényi relative entropy
Converges to relative entropy:
D(pX kqX ) = lim Dα (pX kqX )

α→1
Data-processing inequality for all α ∈ (0, 1) ∪ (1, ∞):
Dα (pX kqX ) ≥ Dα (NY |X (pX )kNY |X (qX ))
Additivity: for probability distributions pX1 and pX2 , measures qX1

and qX2 , and for all α ∈ (0, 1) ∪ (1, ∞):
Dα (pX1 ⊗ pX2 kqX1 ⊗ qX2 ) = Dα (pX1 kqX1 ) + Dα (pX2 kqX2 )
Ordering: For α > β > 0
Dα (pX kqX ) ≥ Dβ (pX kqX )

Quantum states
A quantum state ρA is a positive semi-definite, unit trace operator

(i.e., Hermitian matrix with all eigenvalues non-negative and summing
to one)
Subscript notation indicates ρA is a state of a quantum system A
Also called density operator or density matrix
D(HA ) denotes set of density operators acting on a Hilbert space HA
Classical probability distributions are a special case in which density

operator is diagonal

Quantum channels
A quantum channel NA→B is a completely positive, trace-preserving

map, which evolves system A to system B
Complete positivity and trace preservation are equivalent to Choi state
ΦN
RB := (idR ⊗NA→B )(ΦRA )
being positive semi-definite and TrB [ΦN

RB ] =
1
dR IR , where R ' A and
maximally entangled state ΦRA is defined as
1 X
ΦRA := |iihj|R ⊗ |iihj|A ,
dR
i,j
A classical channel is a special case in which the Choi state is a

diagonal density operator.

Quantum relative entropy
Quantum relative entropy of a state ρ and a positive semi-definite

operator σ is defined as [Ume62]
D(ρkσ) := Tr[ρ(log2 ρ − log2 σ)]
Standard definition with operational meaning [HP91, ON00]

Special cases of quantum relative entropy
Quantum entropy [vN27] is a special case of relative entropy:
H(ρ) := − Tr[ρ log2 ρ] = −D(ρkI )
Quantum mutual information [Str65] is a special case of relative

entropy:
I (A; B)ρ := D(ρAB kρA ⊗ ρB ) = inf D(ρAB kρA ⊗ σB )

σB ∈D(HB )

Data-processing inequality for quantum relative entropy
Let ρ be a state, σ a positive semi-definite operator, and N a quantum

channel. Then [Lin75]
D(ρkσ) ≥ D(N (ρ)kN (σ))

Petz–Rényi relative entropy
There are at least two meaningful ways to generalize the classical

Rényi relative entropy to the quantum case. Let us begin with the
Petz–Rényi relative entropy.
Petz–Rényi relative entropy [Pet86] defined for α ∈ (0, 1) ∪ (1, ∞):
1
Dα (ρkσ) := log2 Tr[ρα σ 1−α ]
α−1

Properties of Petz–Rényi relative entropy
Converges to quantum relative entropy in limit α → 1:
D(ρkσ) = lim Dα (ρkσ)

α→1
Data-processing inequality [Pet86] for all α ∈ (0, 1) ∪ (1, 2]:
Dα (ρkσ) ≥ Dα (N (ρ)kN (σ))
Additivity: For states ρ1 and ρ2 , and positive semi-definite operators

σ1 and σ2 :
Dα (ρ1 ⊗ ρ2 kσ1 ⊗ σ2 ) = Dα (ρ1 kσ1 ) + Dα (ρ2 kσ2 )
Ordering [MH11]: For α > β > 0
Dα (ρkσ) ≥ Dβ (ρkσ)

Sandwiched Rényi relative entropy
Different quantum generalization of classical Rényi relative entropy:
Sandwiched Rényi relative entropy [MLDS+ 13, WWY14] defined

for all α ∈ (0, 1) ∪ (1, ∞):
1
eα (ρkσ) :=
D log2 Tr[(σ (1−α)/2α ρσ (1−α)/2α )α ]
α−1

Properties of sandwiched Rényi relative entropy
Converges to quantum relative entropy in limit α → 1
[MLDS+ 13, WWY14]:
D(ρkσ) = lim D
eα (ρkσ)
α→1
Data-processing inequality for all α ∈ [1/2, 1) ∪ (1, ∞) [FL13]:

eα (ρkσ) ≥ D
D eα (N (ρ)kN (σ))

σ1 and σ2 :
eα (ρ1 ⊗ ρ2 kσ1 ⊗ σ2 ) = D
D eα (ρ1 kσ1 ) + D
eα (ρ2 kσ2 )
Ordering [MLDS+ 13]: For α > β > 0,

eα (ρkσ) ≥ D
D eβ (ρkσ)

Quantum hypothesis testing [Hel67, Hel69, Hol72, Hol73]
Quantum system prepared in the state ρ or σ and objective is to

figure out which one was prepared
Make a quantum measurement {Λ, I − Λ} to figure out which was

prepared
Assign outcome Λ to “guess ρ”, and the outcome I − Λ to “guess σ”
Probability of committing a Type I error (“false alarm”):
Tr[(I − Λ)ρ]
Probability of committing a Type II error (“missed detection”):
Tr[Λσ]

Hypothesis testing relative entropy [BD10, BD11, WR12]
Minimize Type II error probability subject to a constraint on Type I

error probability:
DHε (ρkσ) := − log2 inf {Tr[Λσ] : Tr[(I − Λ)ρ] ≤ ε, Λ ≤ I }

Λ≥0
More generally can define this quantity when σ is positive

semi-definite
Obeys data-processing inequality (operational argument):
DHε (ρkσ) ≥ DHε (N (ρ)kN (σ))

Quantum Stein’s lemma [HP91, ON00]
Optimal achievable rate for hypothesis testing:

1
E (ρ, σ) := inf lim inf DHε (ρ⊗n kσ ⊗n )
ε∈(0,1) n→∞ n
Optimal strong converse rate for hypothesis testing:

1
Ee(ρ, σ) := sup lim sup DHε (ρ⊗n kσ ⊗n )
ε∈(0,1) n→∞ n
Always have E (ρ, σ) ≤ Ee(ρ, σ)
Quantum Stein’s lemma:
E (ρ, σ) = Ee(ρ, σ) = D(ρkσ)

Relating quantum hypothesis testing and relative entropy
Let ρ be a state and σ a positive semi-definite operator
Lower bound [Hay07, AMV12, QWW18]: For ε ∈ (0, 1], and

α ∈ (0, 1):

α 1
DHε (ρkσ) ≥ Dα (ρkσ) + log2
α−1 ε
Upper bound [CMW16]: For ε ∈ [0, 1), and α ∈ (1, ∞):

ε α 1
DH (ρkσ) ≤ D
eα (ρkσ) + log2
α−1 1−ε

Application of lower bound
Apply lower bound and additivity to find for all ε ∈ (0, 1] and
α ∈ (0, 1) that

1 ε ⊗n ⊗n 1 ⊗n ⊗n α 1
DH (ρ kσ ) ≥ Dα (ρ kσ ) + log2
n n n(α − 1) ε

α 1
= Dα (ρkσ) + log2
n(α − 1) ε
Take n → ∞ limit to find for all α ∈ (0, 1) that

1
lim inf DHε (ρ⊗n kσ ⊗n ) ≥ Dα (ρkσ)
n→∞ n
Since lower bound holds for all α ∈ (0, 1), conclude that
1
lim inf DHε (ρ⊗n kσ ⊗n ) ≥ sup Dα (ρkσ) = D(ρkσ)
n→∞ n α∈(0,1)

Application of upper bound
Apply upper bound and additivity to find for all ε ∈ [0, 1) and
α ∈ (1, ∞) that

1 ε ⊗n ⊗n 1 e ⊗n ⊗n α 1
D (ρ kσ ) ≤ Dα (ρ kσ ) + log2
n H n n(α − 1) 1−ε

α 1
=D eα (ρkσ) + log2
n(α − 1) 1−ε
Take n → ∞ limit to find for all α ∈ (1, ∞) that

1
lim sup DHε (ρ⊗n kσ ⊗n ) ≤ D
eα (ρkσ)
n→∞ n
Since upper bound holds for all α ∈ (1, ∞), conclude that
1
lim sup DHε (ρ⊗n kσ ⊗n ) ≤ inf D eα (ρkσ) = D(ρkσ)
n→∞ n α∈(1,∞)

Quantum Stein’s lemma
Combining lower and upper bound gives quantum Stein’s lemma:
E (ρ, σ) = Ee(ρ, σ) = D(ρkσ)
Actually something slightly stronger: for all ε ∈ (0, 1),
1 ε ⊗n ⊗n
lim D (ρ kσ ) = D(ρkσ)
n→∞ n H
Operational interpretation of quantum relative entropy

Generalized divergence
Generalized divergence D(ρkσ) [PV10, SW12] is a function that

satisfies data processing; i.e., for every state ρ, positive semi-definite
operator σ, and channel N , the following inequality holds
D(ρkσ) ≥ D(N (ρ)kN (σ))
Examples include relative entropy, Petz– and sandwiched Rényi

relative entropies (for certain α values), hypothesis testing relative
entropy, etc.

Information measures from generalized divergence
Generalized mutual information of a bipartite state ρAB :
I (A; B)ρ := inf D(ρAB kρA ⊗ σB )

σB ∈D(HB )
Alt. generalized mutual information of a bipartite state ρAB :
Ī (A; B)ρ := D(ρAB kρA ⊗ ρB )

Entanglement measures from generalized divergence
Generalized divergence of entanglement of a bipartite state ρAB
[VP98, Das18]:
E R (A; B) := inf D(ρAB kσAB )

σAB ∈SEP(A:B)
Generalized Rains divergence [TWW17] of a bipartite state ρAB :
R(A; B) := inf D(ρAB kσAB )

σAB ∈PPT0 (A:B)
where PPT0 (A : B) := {σAB : σAB ≥ 0, kTB (σAB )k1 ≤ 1}

X
TB (·) := |iihj|B (·)|iihj|B
i,j
These are entanglement measures [HHHH09] because they do not

increase under the action of local operations and classical
communication on ρAB
Channel information measures
Let NA→B be a quantum channel.
We can define channel measures by optimizing state measures [Wil17].
For example, generalized mutual information of a channel defined by
I (N ) := sup I (R; B)ω ,

ψRA
where ωRB := NA→B (ψRA )
More generally, if S(A; B)ρ is one of the state measures given

previously, then channel measure is
S(N ) := sup S(R; B)ω .

ψRA

Entanglement-assisted classical communication [BSST99]
Suppose Alice and Bob are connected by a quantum channel NA→B .

An (|M|, ε) entanglement-assisted classical comm. code consists of an
shared state ΨA0 B 0 , and encoding channel EM 0 A0 →A , and a decoding
measurement channel DBB 0 →M̂ such that
1
Φ
M M̂ − (DBB 0 →M̂ ◦ NA→B ◦ EM 0 A0 →A )(ΦMM 0 ⊗ ΨA0 B 0 ) 1 ≤ ε,
2
where
1 X
ΦM M̂ := |mihm|M ⊗ |mihm|M̂ ,
dim(HM ) m
|M| = number of messages.
Note that ΦM M̂ represents a classical state, and the goal is for the
coding scheme to preserve the classical correlations in this state.

Schematic of an entanglement-assisted code
Alice Bob
M’ A B
E N
A’ M̂
D
B’

Entanglement-assisted classical capacity
One-shot entanglement-assisted classical capacity [DH13]:
ε
CEA (N ) := sup {log2 |M| : ∃ (|M|, ε) EA code for N }
ΨA0 B 0 ,E,D,M
Entanglement-assisted capacity:
1 ε
CEA (N ) := inf lim inf CEA (N ⊗n )
ε∈(0,1) n→∞ n
Strong converse entanglement-assisted capacity:

1 ε
CeEA (N ) := sup lim sup CEA (N ⊗n )
ε∈(0,1) n→∞ n
Always have CEA (N ) ≤ CeEA (N )

Bounds on one-shot EA capacity
Using methods called position-based coding [AJW19] and

sequential decoding [GLM12, Sen11, OMW19], we find a lower
bound on one-shot EA capacity, holding for η ∈ (0, ε):

¯ε−η 4ε ε
IH (N ) − log2 2 ≤ CEA (N )
η
By relating EA communication task to hypothesis testing, we obtain

an upper bound on one-shot EA capacity [MW14]:
ε
CEA (N ) ≤ IHε (N )

Lower bound on EA capacity
Lower bound on one-shot EA capacity implies lower bound on EA

capacity
Now pick η = ε/2 and, for α ∈ (0, 1), apply lower bound for
hypothesis testing relative entropy from before:

1 ε 1 ε 1 16
CEA (N ⊗n ) ≥ I¯H2 (N ⊗n ) − log2
n n n ε

1¯ ⊗n α 2 1 16
≥ Iα (N ) − log2 − log2
n n(1 − α) ε n ε

α 2 1 16
≥ I¯α (N ) − log2 − log2
n(1 − α) ε n ε

Lower bound on EA capacity (ctd.)
Take n → ∞ limit to find for all α ∈ (0, 1) that
1 ε
lim inf CEA (N ⊗n ) ≥ I¯α (N )
n→∞ n
Since it holds for all α ∈ (0, 1), we conclude that
1 ε
lim inf CEA (N ⊗n ) ≥ sup I¯α (N ) = I (N )
n→∞ n α∈(0,1)
and thus
CEA (N ) ≥ I (N )

Upper bound on EA capacity
Upper bound on one-shot EA capacity implies upper bound on EA

capacity
Apply upper bound for hypothesis testing relative entropy from before
for α ∈ (1, ∞) and additivity of sandwiched Rényi channel mutual
information [DJKR06, GW15]:
1 ε 1
CEA (N ⊗n ) ≤ IHε (N ⊗n )
n n
1 α 1
≤ Ieα (N ⊗n ) + log2
n n(α − 1) 1−ε

α 1
= Iα (N ) +
e log2
n(α − 1) 1−ε

Upper bound on EA capacity (ctd.)
Take n → ∞ limit to find for all α ∈ (1, ∞) that
1 ε
lim sup CEA (N ⊗n ) ≤ Ieα (N )
n→∞ n
Since it holds for all α ∈ (1, ∞), we conclude that
1 ε
lim sup CEA (N ⊗n ) ≤ inf Ieα (N ) = I (N )
n→∞ n α∈(1,∞)
and thus
CeEA (N ) ≤ I (N )

Entanglement-assisted capacity theorem
Combining lower and upper bounds:
CEA (N ) = CeEA (N ) = I (N )
Operational meaning for mutual information of a quantum channel

as entanglement-assisted classical capacity
[BSST02, BCR11, BDH+ 14]

Geometric Rényi relative entropy
Geometric Rényi relative entropy [PR98, Mat13] is a

generalization of classical Rényi relative entropy that is useful for
bounding feedback-assisted capacities [FF19].
For ρ a state, σ a positive semi-definite operator, and

α ∈ (0, 1) ∪ (1, ∞), geometric Rényi relative entropy defined as
1
bα (ρkσ) :=
D log2 Tr[σ(σ −1/2 ρσ −1/2 )α ]
α−1

Namesake for geometric Rényi relative entropy
Called geometric Rényi relative entropy because it can be written in

terms of weighted operator geometric mean [LL01]:
Gα (σ, ρ) := σ 1/2 (σ −1/2 ρσ −1/2 )α σ 1/2
so that
1
D
bα (ρkσ) = log2 Tr[Gα (σ, ρ)]
α−1

Properties of geometric Rényi relative entropy
Converges to Belavkin–Staszewski relative entropy [BS82] in limit
α → 1:
D(ρkσ)
b = lim Dbα (ρkσ)
α→1
where D(ρkσ)
b := Tr[ρ log2 ρ1/2 σ −1 ρ1/2 ]
Data-processing inequality [PR98, Mat13] for α ∈ (0, 1) ∪ (1, 2]:

bα (ρkσ) ≥ D
D bα (N (ρ)kN (σ))

σ1 and σ2 :
bα (ρ1 ⊗ ρ2 kσ1 ⊗ σ2 ) = D
D bα (ρ1 kσ1 ) + D
bα (ρ2 kσ2 )
Ordering [KW20]: For α > β > 0

bα (ρkσ) ≥ D
D bβ (ρkσ)

Geometric Rényi channel divergence [LKDW18, FF19, KW20]
Geometric Rényi channel divergence: Distinguishability measure for

quantum channel NA→B and a completely positive map MA→B , for
α ∈ (0, 1) ∪ (1, 2]:
bα (N kM) := sup D(N
D b A→B (ψRA )kMA→B (ψRA ))
ψRA

Subadditivity of geometric Rényi channel divergence
Key property: subadditivity with respect to serial composition

[FF19].
1
For channels NA→B 2
and NB→C , completely positive maps M1A→B
2
and MB→C , and α ∈ (0, 1) ∪ (1, 2]:
bα (N 2 ◦ N 1 kM2 ◦ M1 ) ≤ D
D bα (N 1 kM1 ) + D
bα (N 2 kM2 )

Geometric Rains entanglement [BDWW19, FF19]
Let NAB→A0 B 0 be a bipartite quantum channel.
Geometric Rains entanglement of NAB→A0 B 0 defined as
Rbα (N ) := inf bα (N kM)

D
M:EN (M)≤0
where logarithmic negativity of MAB→A0 B 0 defined as
EN (M) := log2 kTB0 ◦ MAB→A0 B 0 ◦ TB k
Diamond norm of a Hermiticity-preserving map PC →D defined as
kPk := sup kPC →D (ψRC )k1

ψRC
where ψRC is a pure state with R ' C

Special cases of geometric Rains entanglement
Can be evaluated for a bipartite state ρA0 B 0 , which is a bipartite

channel with AB inputs trivial
Can be evaluated for a point-to-point channel NA→B 0 , which is a

bipartite channel with input B and output A0 trivial

Subadditivity of geometric Rains entanglement [FF19]
Important property of logarithmic negativity: subadditivity with

respect to serial composition.
For completely positive maps M1AB→A0 B 0 and M2A0 B 0 →A00 B 00 :
EN (M2 ◦ M1 ) ≤ EN (M1 ) + EN (M2 )
This and subadditivity of geometric Rényi channel divergence imply

subadditivity for geometric Rényi entanglement: For bipartite
1
channels NAB→A 2
0 B 0 and NA0 B 0 →A00 B 00 and α ∈ (0, 1) ∪ (1, 2]:
Rbα (N 2 ◦ N 1 ) ≤ Rbα (N 1 ) + Rbα (N 2 )

LOCC channel
In the theory of entanglement and quantum communication, one often

assumes that Alice and Bob can communicate classical data for free.
Paradigm is local op.’s and classical comm. (LOCC) [BDSW96].
A one-way LOCC channel from Alice to Bob consists of Alice

performing a quantum instrument, sending classical outcome to Bob,
who performs a quantum channel conditioned on the classical data.
An LOCC channel consists of finite, but arbitrarily large number of

1-way LOCC channels from Alice to Bob and then from Bob to Alice.

LOCC channel (ctd.)
An LOCC channel can be written as a separable channel LAB→A0 B 0 :

X
z z
LAB→A0 B 0 (ρAB ) = (EA→A 0 ⊗ FB→B 0 )(ρAB ),
z
where {EA→A z
0 }z and {FB→B 0 }z are sets of completely positive, trace
non-increasing maps, such that LAB→A0 B 0 is a completely positive,

trace-preserving map (quantum channel).

LOCC-assisted quantum communication
A1’ A2’ A’n MA
A1 B1 A2 B2 An Bn
LOCC N LOCC N LOCC LOCC N LOCC
B1’ B2’ B’n MB
An (n, M, ε) protocol for LOCC-assisted quantum communication

over the quantum channel N calls the channel n times.
In between every channel use, Alice and Bob are allowed to perform
an LOCC channel for free.
The final state ωMA MB should have fidelity larger than 1 − ε with a
maximally entangled state ΦMA MB of Schmidt rank M:
hΦ|MA MB ωMA MB |ΦiMA MB ≥ 1 − ε.

LOCC assisted quantum capacity [BDSW96, TGW14]
n-shot LOCC assisted quantum capacity:
n,ε
Q↔ (N ) := sup {log2 M : ∃ (n, M, ε) protocol for N }
LOCC protocols
LOCC-assisted quantum capacity of N :

1 n,ε
Q↔ (N ) := inf lim inf Q↔ (N )
ε∈(0,1) n→∞ n
Strong converse LOCC-assisted quantum capacity of N defined as
e↔ (N ) := sup lim sup 1 Q n,ε (N )

Q ↔
ε∈(0,1) n→∞ n
Always have Q↔ (N ) ≤ Q
e↔ (N )
Upper bound on n-shot LOCC-a. quantum capacity [FF19]
For final state ωMA MB , can show for all ε ∈ [0, 1) and α ∈ (1, ∞) that
log2 M ≤ RHε (MA ; MB )ω

α 1
≤ Rbα (MA ; MB )ω + log2
α−1 1−ε

α 1
≤ nRbα (N ) + log2
α−1 1−ε
It then follows for all α ∈ (1, ∞) that

1 n,ε α 1
Q (N ) ≤ Rα (N ) +
b log2
n ↔ n(α − 1) 1−ε

Upper bound on LOCC-assisted quantum capacity
Now take the limit n → ∞:

1 n,ε
lim sup Q↔ (N ) ≤ Rbα (N )
n→∞ n
Since the bound holds for all α ∈ (1, ∞), conclude that
1 n,ε
lim sup Q↔ (N ) ≤ inf Rbα (N ) = R(N
b )
n→∞ n α∈(1,∞)
Conclude bound on strong converse LOCC-assisted quantum capacity

e↔ (N ) ≤ R(N
Q b )

Conclusion
Three quantum generalizations of Rényi relative entropy and their use:

Petz–Rényi relative entropy:
1
Dα (ρkσ) := log2 Tr[ρα σ 1−α ]
α−1
Finds use as lower bound for distinguishability and comm. tasks
Sandwiched Rényi relative entropy:

1
eα (ρkσ) :=
D log2 Tr[(σ (1−α)/2α ρσ (1−α)/2α )α ]
α−1
Finds use as upper bound for distinguishability and comm. tasks

Conclusion (ctd.)
Geometric Rényi relative entropy:

1
bα (ρkσ) :=
D log2 Tr[σ(σ −1/2 ρσ −1/2 )α ]
α−1
Finds use as upper bound for feedback-assisted distinguishability and
communication tasks

Future directions
Are there other interesting quantum generalizations of Rényi relative

entropy?
Do they have applications in quantum information theory?
See α-z Rényi relative entropies [AD15] and their data-processing

inequality [Zha20]

References I
[AD15] Koenraad M. R. Audenaert and Nilanjana Datta. α-z-Rényi relative

entropies. Journal of Mathematical Physics, 56(2):022202, February 2015.
arXiv:1310.7178.
[AJW19] Anurag Anshu, Rahul Jain, and Naqueeb A. Warsi. Building blocks for
communication over noisy quantum networks. IEEE Transactions on
Information Theory, 65(2):1287–1306, February 2019. arXiv:1702.01940.
[AMV12] Koenraad M. R. Audenaert, Milan Mosonyi, and Frank Verstraete.

Quantum state discrimination bounds for finite sample size. Journal of
Mathematical Physics, 53(12):122205, December 2012. arXiv:1204.0711.
[BCR11] Mario Berta, Matthias Christandl, and Renato Renner. The quantum
reverse Shannon theorem based on one-shot information theory.
Communications in Mathematical Physics, 306(3):579–615, August 2011.
arXiv:0912.3805.
[BD10] Francesco Buscemi and Nilanjana Datta. The quantum capacity of
channels with arbitrarily correlated noise. IEEE Transactions on Information
Theory, 56(3):1447–1460, March 2010. arXiv:0902.0158.

References II
[BD11] Fernando G. S. L. Brandao and Nilanjana Datta. One-shot rates for

entanglement manipulation under non-entangling maps. IEEE Transactions
on Information Theory, 57(3):1754–1760, March 2011. arXiv:0905.2673.
[BDH+ 14] Charles H. Bennett, Igor Devetak, Aram W. Harrow, Peter W. Shor, and
Andreas Winter. The quantum reverse Shannon theorem and resource
tradeoffs for simulating quantum channels. IEEE Transactions on
Information Theory, 60(5):2926–2959, May 2014. arXiv:0912.5537.
[BDSW96] Charles H. Bennett, David P. DiVincenzo, John A. Smolin, and William K.

Wootters. Mixed-state entanglement and quantum error correction.
Physical Review A, 54(5):3824–3851, November 1996.
arXiv:quant-ph/9604024.
[BDWW19] Stefan Bäuml, Siddhartha Das, Xin Wang, and Mark M. Wilde. Resource
theory of entanglement for bipartite quantum channels. July 2019.
arXiv:1907.04181.

References III
[BS82] V. P. Belavkin and P. Staszewski. C*-algebraic generalization of relative

entropy and entropy. Annales de l’I.H.P. Physique théorique, 37(1):51–58,
1982.
[BSST99] Charles H. Bennett, Peter W. Shor, John A. Smolin, and Ashish V.
Thapliyal. Entanglement-assisted classical capacity of noisy quantum
channels. Physical Review Letters, 83(15):3081–3084, October 1999.
[BSST02] Charles H. Bennett, Peter W. Shor, John A. Smolin, and Ashish V.

Thapliyal. Entanglement-assisted capacity of a quantum channel and the
reverse Shannon theorem. IEEE Transactions on Information Theory,
48(10):2637–2655, October 2002. arXiv:quant-ph/0106052.
[CMW16] Tom Cooney, Milan Mosonyi, and Mark M. Wilde. Strong converse
exponents for a quantum channel discrimination problem and
quantum-feedback-assisted communication. Communications in
Mathematical Physics, 344(3):797–829, June 2016. arXiv:1408.3373.

References IV
[Csi95] Imre Csiszár. Generalized cutoff rates and Rényi’s information measures.
IEEE Transactions on Information Theory, 41(1):26–34, January 1995.
[Das18] Siddhartha Das. Bipartite Quantum Interactions: Entangling and

Information Processing Abilities. PhD thesis, Louisiana State University,
October 2018. Available at
https://digitalcommons.lsu.edu/gradschool_dissertations/4717/
and arXiv:1901.05895.
[DH13] Nilanjana Datta and Min-Hsiu Hsieh. One-shot entanglement-assisted
quantum and classical communication. IEEE Transactions on Information
Theory, 59(3):1929–1939, March 2013. arXiv:1105.3321.
[DJKR06] Igor Devetak, Marius Junge, Christopher King, and Mary Beth Ruskai.
Multiplicativity of completely bounded p-norms implies a new additivity
result. Communications in Mathematical Physics, 266(1):37–63, August
2006. arXiv:quant-ph/0506196.

References V
[FF19] Kun Fang and Hamza Fawzi. Geometric Rényi divergence and its
applications in quantum channel capacities. September 2019.
arXiv:1909.05758v1.
[FL13] Rupert L. Frank and Elliott H. Lieb. Monotonicity of a relative Rényi
entropy. Journal of Mathematical Physics, 54(12):122201, December 2013.
arXiv:1306.5358.
[GLM12] Vittorio Giovannetti, Seth Lloyd, and Lorenzo Maccone. Achieving the
Holevo bound via sequential measurements. Physical Review A,
85(1):012302, January 2012. arXiv:1012.0386.
[GW15] Manish Gupta and Mark M. Wilde. Multiplicativity of completely bounded

p-norms implies a strong converse for entanglement-assisted capacity.
Communications in Mathematical Physics, 334(2):867–887, March 2015.
arXiv:1310.7028.
[Hay07] Masahito Hayashi. Error exponent in asymmetric quantum hypothesis
testing and its application to classical-quantum channel coding. Physical
Review A, 76(6):062301, December 2007. arXiv:quant-ph/0611013.

References VI
[Hel67] Carl W. Helstrom. Detection theory and quantum mechanics. Information
and Control, 10(3):254–291, 1967.
[Hel69] Carl W. Helstrom. Quantum detection and estimation theory. Journal of

Statistical Physics, 1:231–252, 1969.
[HHHH09] Ryszard Horodecki, Pawel Horodecki, Michal Horodecki, and Karol

Horodecki. Quantum entanglement. Reviews of Modern Physics,
81(2):865–942, June 2009. arXiv:quant-ph/0702225.
[Hol72] Alexander S. Holevo. An analogue of statistical decision theory and

noncommutative probability theory. Trudy Moskovskogo Matematicheskogo
Obshchestva, 26:133–149, 1972.
[Hol73] Alexander S. Holevo. Statistical problems in quantum physics. In Second

Japan-USSR Symposium on Probability Theory, volume 330 of Lecture
Notes in Mathematics, pages 104–119. Springer Berlin / Heidelberg, 1973.
[HP91] Fumio Hiai and Dénes Petz. The proper formula for relative entropy and its
asymptotics in quantum probability. Communications in Mathematical
Physics, 143(1):99–114, December 1991.

References VII
[KL51] S. Kullback and R. A. Leibler. On information and sufficiency. The Annals

of Mathematical Statistics, 22(1):79–86, March 1951.
[KW20] Vishal Katariya and Mark M. Wilde. Geometric distinguishability measures

limit quantum channel estimation and discrimination. April 2020.
arXiv:2004.10708.
[Lin75] Göran Lindblad. Completely positive maps and entropy inequalities.
Communications in Mathematical Physics, 40(2):147–151, June 1975.
[LKDW18] Felix Leditzky, Eneet Kaur, Nilanjana Datta, and Mark M. Wilde.
Approaches for approximate additivity of the Holevo information of
quantum channels. Physical Review A, 97(1):012332, January 2018.
arXiv:1709.01111.
[LL01] Jimmie D. Lawson and Yongdo Lim. The geometric mean, matrices,
metrics, and more. The American Mathematical Monthly, 108(9):797–812,
November 2001.
[Mat13] Keiji Matsumoto. A new quantum version of f -divergence. 2013.
arXiv:1311.4722.

References VIII
[MH11] Milán Mosonyi and Fumio Hiai. On the quantum Rényi relative entropies
and related capacity formulas. IEEE Transactions on Information Theory,
57(4):2474–2487, April 2011. arXiv:0912.1286.
[MLDS+ 13] Martin Müller-Lennert, Frédéric Dupuis, Oleg Szehr, Serge Fehr, and
Marco Tomamichel. On quantum Rényi entropies: a new generalization
and some properties. Journal of Mathematical Physics, 54(12):122203,
December 2013. arXiv:1306.3142.
[MW14] William Matthews and Stephanie Wehner. Finite blocklength converse
bounds for quantum channels. IEEE Transactions on Information Theory,
60(11):7317–7329, November 2014. arXiv:1210.4722.
[OMW19] Samad Khabbazi Oskouei, Stefano Mancini, and Mark M. Wilde. Union
bound for quantum information processing. Proceedings of the Royal
Society A, 475(2221):20180612, January 2019. arXiv:1804.08144.
[ON00] Tomohiro Ogawa and Hiroshi Nagaoka. Strong converse and Stein’s lemma
in quantum hypothesis testing. IEEE Transactions on Information Theory,
46(7):2428–2433, November 2000. arXiv:quant-ph/9906090.

References IX
[Pet86] Dénes Petz. Quasi-entropies for finite quantum systems. Reports in

Mathematical Physics, 23:57–65, 1986.
[PR98] Dénes Petz and Mary Beth Ruskai. Contraction of generalized relative
entropy under stochastic mappings on matrices. Infinite Dimensional
Analysis, Quantum Probability and Related Topics, 1(1):83–89, January
1998.
[PV10] Yury Polyanskiy and Sergio Verdú. Arimoto channel coding converse and
Rényi divergence. In Proceedings of the 48th Annual Allerton Conference
on Communication, Control, and Computation, pages 1327–1333,
September 2010.
[QWW18] Haoyu Qi, Qing-Le Wang, and Mark M. Wilde. Applications of

position-based coding to classical communication over quantum channels.
Journal of Physics A, 51(44):444002, November 2018. arXiv:1704.01361.

References X
[Rén61] Alfréd Rényi. On measures of entropy and information. Proceedings of the

4th Berkeley Symposium on Mathematics, Statistics and Probability,
1:547–561, 1961. held at the Statistical Laboratory, University of California,
1960, edited by J. Neyman (University of California Press, Berkeley).
[Sen11] Pranab Sen. Achieving the Han-Kobayashi inner bound for the quantum
interference channel by sequential decoding. September 2011.
arXiv:1109.0802.
[Sha48] Claude E. Shannon. A mathematical theory of communication. Bell System
Technical Journal, 27:379–423, 1948.
[Str65] Ruslan L. Stratonovich. Information capacity of a quantum

communications channel. i. Soviet Radiophysics, 8(1):82–91, January 1965.
[SW12] Naresh Sharma and Naqueeb A. Warsi. On the strong converses for the
quantum channel capacity theorems. May 2012. arXiv:1205.1712.
[TGW14] Masahiro Takeoka, Saikat Guha, and Mark M. Wilde. The squashed
entanglement of a quantum channel. IEEE Transactions on Information
Theory, 60(8):4987–4998, August 2014. arXiv:1310.0129.

References XI
[TWW17] Marco Tomamichel, Mark M. Wilde, and Andreas Winter. Strong converse
rates for quantum communication. IEEE Transactions on Information
Theory, 63(1):715–727, January 2017. arXiv:1406.2946.
[Ume62] Hisaharu Umegaki. Conditional expectations in an operator algebra IV

(entropy and information). Kodai Mathematical Seminar Reports,
14(2):59–85, 1962.
[vN27] Johann von Neumann. Thermodynamik quantenmechanischer

gesamtheiten. Nachrichten von der Gesellschaft der Wissenschaften zu
Göttingen, Mathematisch-Physikalische Klasse, 102:273–291, 1927.
[VP98] Vlatko Vedral and Martin B. Plenio. Entanglement measures and

purification procedures. Physical Review A, 57(3):1619–1633, March 1998.
[Wil17] Mark M. Wilde. Quantum Information Theory. Cambridge University Press,

second edition, 2017. arXiv:1106.1445v7.

References XII
[WR12] Ligong Wang and Renato Renner. One-shot classical-quantum capacity and
hypothesis testing. Physical Review Letters, 108(20):200501, 2012.
arXiv:1007.5456.
[WWY14] Mark M. Wilde, Andreas Winter, and Dong Yang. Strong converse for the
classical capacity of entanglement-breaking and Hadamard channels via a
sandwiched Rényi relative entropy. Communications in Mathematical
Physics, 331(2):593–622, October 2014. arXiv:1306.1586.
[Zha20] Haonan Zhang. From wigner-yanase-dyson conjecture to carlen-frank-lieb

conjecture. Advances in Mathematics, 365:107053, May 2020.
arXiv:1811.01205.

On The Quantum Renyi Relative Entropies and Their Use

Uploaded by

Copyright:

Available Formats

On The Quantum Renyi Relative Entropies and Their Use

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

On The Quantum Renyi Relative Entropies and Their Use

Uploaded by

Copyright:

Available Formats

On the quantum Rényi relative entropies and their use

Hearne Institute for Theoretical Physics,

ISL Colloquium, Stanford University, Stanford, California, September 17, 2020

Mark M. Wilde (LSU) 1 / 68

Denes Petz, “Quantum entropy and its use” (1991)

Denes Petz, “From f -divergence to quantum quasi-entropies and their

Felix Leditzky, “Relative entropies and their use in quantum

M. E. Shirokov, “Upper bounds for the Holevo quantity and their

Mark M. Wilde (LSU) 2 / 68

Entropy and relative entropy are at the core of information theory

Rényi entropy and Rényi relative entropy are fruitful

When generalizing to the quantum case, there is not a unique way to

In this talk, I’ll review three quantum generalizations of Rényi relative

I’ll also discuss their applications in quantum information theory

Mark M. Wilde (LSU) 3 / 68

Figure: Screen capture from the movie Arrival (2016)

Equal to the optimal limit of classical data compression [Sha48]

Relative entropy [KL51] generalizes entropy

Let pX ∈ P(X ) be a probability distribution, and let qX be a

Distinguishability measure with interpretation in hypothesis testing

Mark M. Wilde (LSU) 5 / 68

Entropy is a special case of relative entropy:

H(pX ) = −D(pX k1),

where 1 is the vector of all ones.

Mutual information is a special case of relative entropy:

I (X ; Y ) := D(pXY kpX ⊗ pY ) = inf D(pXY kpX ⊗ qY )

Mark M. Wilde (LSU) 6 / 68

Let NY |X be a classical channel and define

Then the data-processing inequality for relative entropy is

D(pX kqX ) ≥ D(NY |X (pX )kNY |X (qX ))

Mark M. Wilde (LSU) 7 / 68

Rényi relative entropy is a generalization of relative entropy with

Mark M. Wilde (LSU) 8 / 68

Rényi entropy [Rén61] is a special case:

Rényi mutual information [Csi95] defined as

Iα (X ; Y ) := inf Dα (pXY kpX ⊗ qY )

Mark M. Wilde (LSU) 9 / 68

D(pX kqX ) = lim Dα (pX kqX )

Data-processing inequality for all α ∈ (0, 1) ∪ (1, ∞):

Dα (pX kqX ) ≥ Dα (NY |X (pX )kNY |X (qX ))

Additivity: for probability distributions pX1 and pX2 , measures qX1

Dα (pX1 ⊗ pX2 kqX1 ⊗ qX2 ) = Dα (pX1 kqX1 ) + Dα (pX2 kqX2 )

Ordering: For α > β > 0

Dα (pX kqX ) ≥ Dβ (pX kqX )

Mark M. Wilde (LSU) 10 / 68

A quantum state ρA is a positive semi-definite, unit trace operator

Subscript notation indicates ρA is a state of a quantum system A

Also called density operator or density matrix

D(HA ) denotes set of density operators acting on a Hilbert space HA

Classical probability distributions are a special case in which density

Mark M. Wilde (LSU) 11 / 68

A quantum channel NA→B is a completely positive, trace-preserving

Complete positivity and trace preservation are equivalent to Choi state

being positive semi-definite and TrB [ΦN

A classical channel is a special case in which the Choi state is a

Mark M. Wilde (LSU) 12 / 68

Quantum relative entropy of a state ρ and a positive semi-definite

D(ρkσ) := Tr[ρ(log2 ρ − log2 σ)]

Standard definition with operational meaning [HP91, ON00]