Download presentation
Presentation is loading. Please wait.
1
Effcient quantum protocols for XOR functions
Shengyu Zhang The Chinese University of Hong Kong
2
Communication complexity
π₯ π¦ πΉ(π₯,π¦) Two parties, Alice and Bob, jointly compute a function π on input (π₯,π¦). π₯ known only to Alice and π¦ only to Bob. Communication complexity*1: how many bits are needed to be exchanged? *1. A. Yao. STOC, 1979.
3
Computation modes Deterministic: Players run determ. protocol. ---π·(πΉ)
Randomized: Players have access to random bits; small error probability allowed. --- π
(πΉ) Quantum: Players send quantum messages. Share resource? (Superscript.) π β (πΉ): share entanglement. π(πΉ): share nothing Error? (Subscript) π π (πΉ): bounded-error. π πΈ (πΉ): zero-error, fixed length.
4
Lower bounds Not only interesting on its own, but also important because of numerous applications. to prove lower bounds. Question: How to lower bound communication complexity itself? Communication matrix π πΉ β πΉ π₯,π¦ π₯,π¦ : π¦ β π₯β πΉ(π₯,π¦)
5
Log-rank conjecture Rank lower bound*1
Conj: The rank lower bound is polynomially tight. combinatorial measure β linear algebra measure. Equivalent to a bunch of other conjectures. related to graph theory*2; nonnegative rank*3, Boolean roots of polynomials*4, quantum sampling complexity*5. β€ log π 1 ππππ π πΉ *1. Melhorn, Schmidt. STOC, 1982. *2. LovΓ‘sz, Saks. FOCS, 1988. *3. LovΓ‘sz. Book Chapter, 1990. *4. Valiant. Info. Proc. Lett., 2004. *5. Ambainis, Schulman, Ta-Shma, Vazirani, Wigderson, SICOMP 2003.
6
Log-rank conjecture: quantum version
Rank lower bound log 2 ππππ π πΉ β€π· πΉ Quantum: rank lower bound *1 log 2 πππ π π π πΉ β€ π π πΉ β€ log π 1 πππ π π π πΉ πππ π π π = min π β² :|π(π₯,π¦)β π β² (π₯,π¦)|β€π ππππ( π β² ) β€ log π 1 ππππ π πΉ *1. Buhrman, de Wolf. CCC, 2001.
7
Log-rank conjecture for XOR functions
Since Log-rank conjecture appears too hard in its full generality,β¦ letβs try some special class of functions. XOR functions: π(π₯βπ¦) πΉ=πββ The linear composition of π₯ and π¦. Include important functions such as Equality, Hamming Distance, Gap Hamming Distance. Interesting connections to Fourier analysis of functions on 0,1 π .
8
Digression: Fourier analysis
βπ: 0,1 π ββ can be written as π= πΌβ 0,1 π π πΌ π πΌ π πΌ π₯ = β1 πΌβ
π₯ , and characters are orthogonal π πΌ :πΌβ 0,1 π : Fourier coefficients of π Parseval: If π
ππππ π = +1,β1 , then πΌ π πΌ 2 =1. Two important measures: π 1 = πΌ π πΌ Spectral norm. π 0 = πΌ: π πΌ β Fourier sparsity. Cauchy-Schwartz: π 1 β€ π 0 1/2 for π: 0,1 π βΒ±1
9
Log-rank Conj. For XOR functions
Interesting connections to Fourier analysis: 1. ππππ π πββ = π 0 . Log-rank Conj: π· πββ β€ log 2 π(1) π Thm.*1 π· πββ β€ 2 π 2 /2 log πβ2 π 1 Thm.*1 π· πββ β€π π π π= deg 2 (π) : degree of π as polynomial over π½ 2 . Fact*2. πβ€ log π *1. Tsang, Wong, Xie, Zhang, FOCS, 2013. *2. Bernasconi and Codenotti. IEEE Transactions on Computers, 1999.
10
Quantum This talk: A simpler case π πΈ πββ =O 2 π log π 0 .
π 1,π = min π: πβπ β β€π π 1 πππ π π π = min π β² :|π(π₯,π¦)β π β² (π₯,π¦)|β€π ππππ( π β² ) This paper: π πββ β€ π 2 π log π 1,π , where π= deg 2 (π) . Recall classical: π· πββ β€ 2 π 2 /2 log πβ2 π 1 Confirms quantum Log-rank Conjecture for low-degree XOR functions. This talk: A simpler case π πΈ πββ =O 2 π log π π πΈ πββ =Ξ© log π *2 Ξ© log πππ π π π πββ β₯ *1. Lee and Shraibman. Foundations and Trends in Theoretical Computer Science, 2009. *2. Buhrman and de Wolf. CCC, 2001.
11
About quantum protocol
Much simpler. log π comes very naturally. Inherently quantum. Not from quantizing any classical protocol.
12
Goal: compute π(π₯+π¦) where π: 0,1 π β Β±1 0 + 1 2 β
β π = πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π Add phase π πΌ (π¦) πβ² = πΌβπ΄ π πΌ π πΌ π₯ π πΌ (π¦) πΈ πΌ πβ² πΈ πΌ β πΌ πΌβπ΄ π πΌ π πΌ π₯+π¦ πΌ Fourier: πΌ β π‘ π πΌ π‘ |π‘βͺ π‘ πΌβπ΄ π πΌ π πΌ π₯+π¦ π πΌ (π‘) π‘ = π‘ π π₯+π¦+π‘ π‘ π‘ π(π₯+π¦+π‘) π‘
13
Goal: compute π(π₯+π¦) where π: 0,1 π β Β±1 π =|+βͺ πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π
π =|+βͺ πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π Add phase π πΌ (π¦) πβ² Decoding + Fourier One more issue: Only Alice knows π‘! Bob doesnβt. Itβs unaffordable to send π‘. Obs: π π’ππ Ξ π‘ π βπ΄+π΄,βπ‘. π΄=π π’ππ π π‘ π(π₯+π¦+π‘) π‘ Measure{|+βͺ,|ββͺ} Measure π‘ A random π‘β 0,1 π and π(π₯+π¦+π‘). Recall our target: π(π₯+π¦). Whatβs the difference? The derivative: Ξ π‘ π π§ =π π§+π‘ π(π§). Good: deg 2 ( Ξ π‘ π) β€ deg 2 (π) β1. Bad: Ξ π‘ π 0 β€ π (Thatβs where the factor of 2 π comes from.)
14
Goal: compute π(π₯+π¦) where π: 0,1 π β Β±1 0,1 π
0,1 π π =|+βͺ πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π Add phase π πΌ (π¦) π΄+π΄ β π+π:π,πβπ΄ πβ² Decoding + Fourier One more issue: Only Alice knows π‘! Bob doesnβt. Itβs unaffordable to send π‘. Obs: π π’ππ Ξ π‘ π βπ΄+π΄,βπ‘. π΄=π π’ππ π π‘ π(π₯+π¦+π‘) π‘ π΄ Measure{|+βͺ,|ββͺ} Measure π‘ A random π‘β 0,1 π and π(π₯+π¦+π‘). Recall our target: π(π₯+π¦). Whatβs the difference? The derivative: Ξ π‘ π π§ =π π§+π‘ π(π§). Good: deg 2 ( Ξ π‘ π) β€ deg 2 (π) β1. Bad: Ξ π‘ π 0 β€ π (Thatβs where the factor of 2 π comes from.)
15
Goal: compute π(π₯+π¦) where π: 0,1 π β Β±1 π =|+βͺ πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π
π =|+βͺ πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π Add phase π πΌ (π¦) πβ² Decoding + Fourier One more issue: Only Alice knows π‘! Bob doesnβt. Itβs unaffordable to send π‘. Obs: π π’ππ Ξ π‘ π βπ΄+π΄,βπ‘. π΄=π π’ππ π Thus in round 2, Alice and Bob can just encode the entire π΄+π΄. π‘ π(π₯+π¦+π‘) π‘ Measure{|+βͺ,|ββͺ} Measure π‘ A random π‘β 0,1 π and π(π₯+π¦+π‘). Recall our target: π(π₯+π¦). Whatβs the difference? The derivative: Ξ π‘ π π§ =π π§+π‘ π(π§). Good: deg 2 ( Ξ π‘ π) β€ deg 2 (π) β1. Bad: Ξ π‘ π 0 β€ π (Thatβs where the factor of 2 π comes from.)
16
At last, deg 2 π π =0, a constant function.
Goal: compute π(π₯+π¦) where π: 0,1 π β Β±1 π =|+βͺ πΌβπ΄ π πΌ π πΌ π₯ πΈ πΌ π Add phase π πΌ (π¦) πβ² Decoding + Fourier π‘ π(π₯+π¦+π‘) π‘ Measure{|+βͺ,|ββͺ} Measure π‘ A random π‘β 0,1 π and π(π₯+π¦+π‘). Compute π 2 β Ξ π‘ π π§ =π π§+π‘ π(π§). At last, deg 2 π π =0, a constant function. Cost: log π΄ + log 2π΄ + log 4π΄ +β¦+ log 2 πβ1 π΄ β€ 2 π log |π΄| . Used trivial bound: ππ΄ β€ π΄ π
17
Open problems Get rid of the factor 2 π !
What can we say about additive structure of supp π for Boolean functions π? Say, π΄+π΄ β€ π΄ 1.5 ?
18
Thanks
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.