Presentation is loading. Please wait.

Presentation is loading. Please wait.

Effcient quantum protocols for XOR functions

Similar presentations


Presentation on theme: "Effcient quantum protocols for XOR functions"β€” Presentation transcript:

1 Effcient quantum protocols for XOR functions
Shengyu Zhang The Chinese University of Hong Kong

2 Communication complexity
π‘₯ 𝑦 𝐹(π‘₯,𝑦) Two parties, Alice and Bob, jointly compute a function 𝑓 on input (π‘₯,𝑦). π‘₯ known only to Alice and 𝑦 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. ---𝐷(𝐹)
Randomized: Players have access to random bits; small error probability allowed. --- 𝑅(𝐹) Quantum: Players send quantum messages. Share resource? (Superscript.) 𝑄 βˆ— (𝐹): share entanglement. 𝑄(𝐹): share nothing Error? (Subscript) 𝑄 πœ– (𝐹): bounded-error. 𝑄 𝐸 (𝐹): 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 𝑀 𝐹 ≝ 𝐹 π‘₯,𝑦 π‘₯,𝑦 : 𝑦 ↓ π‘₯β†’ 𝐹(π‘₯,𝑦)

5 Log-rank conjecture Rank lower bound*1
Conj: The rank lower bound is polynomially tight. combinatorial measure ⇔ 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. ≀ log 𝑂 1 π‘Ÿπ‘Žπ‘›π‘˜ 𝑀 𝐹 *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 π‘Ÿπ‘Žπ‘›π‘˜ 𝑀 𝐹 ≀𝐷 𝐹 Quantum: rank lower bound *1 log 2 π‘Ÿπ‘Žπ‘› π‘˜ πœ– 𝑀 𝐹 ≀ 𝑄 πœ– 𝐹 ≀ log 𝑂 1 π‘Ÿπ‘Žπ‘› π‘˜ πœ– 𝑀 𝐹 π‘Ÿπ‘Žπ‘› π‘˜ πœ– 𝑀 = min 𝑀 β€² :|𝑀(π‘₯,𝑦)βˆ’ 𝑀 β€² (π‘₯,𝑦)|β‰€πœ– π‘Ÿπ‘Žπ‘›π‘˜( 𝑀 β€² ) ≀ log 𝑂 1 π‘Ÿπ‘Žπ‘›π‘˜ 𝑀 𝐹 *1. Buhrman, de Wolf. CCC, 2001.

7 Log-rank conjecture for XOR functions
Since Log-rank conjecture appears too hard in its full generality,… let’s try some special class of functions. XOR functions: 𝑓(π‘₯βŠ•π‘¦) 𝐹=π‘“βˆ˜βŠ• The linear composition of π‘₯ and 𝑦. Include important functions such as Equality, Hamming Distance, Gap Hamming Distance. Interesting connections to Fourier analysis of functions on 0,1 𝑛 .

8 Digression: Fourier analysis
βˆ€π‘“: 0,1 𝑛 →ℝ can be written as 𝑓= π›Όβˆˆ 0,1 𝑛 𝑓 𝛼 πœ’ 𝛼 πœ’ 𝛼 π‘₯ = βˆ’1 𝛼⋅π‘₯ , and characters are orthogonal 𝑓 𝛼 :π›Όβˆˆ 0,1 𝑛 : Fourier coefficients of 𝑓 Parseval: If π‘…π‘Žπ‘›π‘”π‘’ 𝑓 = +1,βˆ’1 , then 𝛼 𝑓 𝛼 2 =1. Two important measures: 𝑓 1 = 𝛼 𝑓 𝛼 Spectral norm. 𝑓 0 = 𝛼: 𝑓 𝛼 β‰  Fourier sparsity. Cauchy-Schwartz: 𝑓 1 ≀ 𝑓 0 1/2 for 𝑓: 0,1 𝑛 β†’Β±1

9 Log-rank Conj. For XOR functions
Interesting connections to Fourier analysis: 1. π‘Ÿπ‘Žπ‘›π‘˜ 𝑀 π‘“βˆ˜βŠ• = 𝑓 0 . Log-rank Conj: 𝐷 π‘“βˆ˜βŠ• ≀ log 2 𝑂(1) 𝑓 Thm.*1 𝐷 π‘“βˆ˜βŠ• ≀ 2 𝑑 2 /2 log π‘‘βˆ’2 𝑓 1 Thm.*1 𝐷 π‘“βˆ˜βŠ• ≀𝑂 𝑑 𝑓 𝑑= deg 2 (𝑓) : degree of 𝑓 as polynomial over 𝔽 2 . Fact*2. 𝑑≀ log 𝑓 *1. Tsang, Wong, Xie, Zhang, FOCS, 2013. *2. Bernasconi and Codenotti. IEEE Transactions on Computers, 1999.

10 Quantum This talk: A simpler case 𝑄 𝐸 π‘“βˆ˜βŠ• =O 2 𝑑 log 𝑓 0 .
𝑓 1,πœ– = min 𝑔: π‘“βˆ’π‘” ∞ β‰€πœ– 𝑔 1 π‘Ÿπ‘Žπ‘› π‘˜ πœ– 𝑀 = min 𝑀 β€² :|𝑀(π‘₯,𝑦)βˆ’ 𝑀 β€² (π‘₯,𝑦)|β‰€πœ– π‘Ÿπ‘Žπ‘›π‘˜( 𝑀 β€² ) This paper: 𝑄 π‘“βˆ˜βŠ• ≀ 𝑂 2 𝑑 log 𝑓 1,πœ– , where 𝑑= deg 2 (𝑓) . Recall classical: 𝐷 π‘“βˆ˜βŠ• ≀ 2 𝑑 2 /2 log π‘‘βˆ’2 𝑓 1 Confirms quantum Log-rank Conjecture for low-degree XOR functions. This talk: A simpler case 𝑄 𝐸 π‘“βˆ˜βŠ• =O 2 𝑑 log 𝑓 𝑄 𝐸 π‘“βˆ˜βŠ• =Ξ© log 𝑓 *2 Ξ© log π‘Ÿπ‘Žπ‘› π‘˜ πœ– 𝑀 π‘“βˆ˜βŠ• 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 𝑓 comes very naturally. Inherently quantum. Not from quantizing any classical protocol.

12 Goal: compute 𝑓(π‘₯+𝑦) where 𝑓: 0,1 𝑛 β†’ Β±1 0 + 1 2 βŠ—
βŠ— πœ“ = π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“ Add phase πœ’ 𝛼 (𝑦) πœ“β€² = π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ πœ’ 𝛼 (𝑦) 𝐸 𝛼 πœ“β€² 𝐸 𝛼 β†’ 𝛼 π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯+𝑦 𝛼 Fourier: 𝛼 β†’ 𝑑 πœ’ 𝛼 𝑑 |𝑑βŒͺ 𝑑 π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯+𝑦 πœ’ 𝛼 (𝑑) 𝑑 = 𝑑 𝑓 π‘₯+𝑦+𝑑 𝑑 𝑑 𝑓(π‘₯+𝑦+𝑑) 𝑑

13 Goal: compute 𝑓(π‘₯+𝑦) where 𝑓: 0,1 𝑛 β†’ Β±1 πœ“ =|+βŒͺ π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“
πœ“ =|+βŒͺ π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“ Add phase πœ’ 𝛼 (𝑦) πœ“β€² Decoding + Fourier One more issue: Only Alice knows 𝑑! Bob doesn’t. It’s unaffordable to send 𝑑. Obs: 𝑠𝑒𝑝𝑝 Ξ” 𝑑 𝑓 βŠ†π΄+𝐴,βˆ€π‘‘. 𝐴=𝑠𝑒𝑝𝑝 𝑓 𝑑 𝑓(π‘₯+𝑦+𝑑) 𝑑 Measure{|+βŒͺ,|βˆ’βŒͺ} Measure 𝑑 A random π‘‘βˆˆ 0,1 𝑛 and 𝑓(π‘₯+𝑦+𝑑). Recall our target: 𝑓(π‘₯+𝑦). What’s the difference? The derivative: Ξ” 𝑑 𝑓 𝑧 =𝑓 𝑧+𝑑 𝑓(𝑧). Good: deg 2 ( Ξ” 𝑑 𝑓) ≀ deg 2 (𝑓) βˆ’1. Bad: Ξ” 𝑑 𝑓 0 ≀ 𝑓 (That’s where the factor of 2 𝑑 comes from.)

14 Goal: compute 𝑓(π‘₯+𝑦) where 𝑓: 0,1 𝑛 β†’ Β±1 0,1 𝑛
0,1 𝑛 πœ“ =|+βŒͺ π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“ Add phase πœ’ 𝛼 (𝑦) 𝐴+𝐴 ≝ π‘Ž+𝑏:π‘Ž,π‘βˆˆπ΄ πœ“β€² Decoding + Fourier One more issue: Only Alice knows 𝑑! Bob doesn’t. It’s unaffordable to send 𝑑. Obs: 𝑠𝑒𝑝𝑝 Ξ” 𝑑 𝑓 βŠ†π΄+𝐴,βˆ€π‘‘. 𝐴=𝑠𝑒𝑝𝑝 𝑓 𝑑 𝑓(π‘₯+𝑦+𝑑) 𝑑 𝐴 Measure{|+βŒͺ,|βˆ’βŒͺ} Measure 𝑑 A random π‘‘βˆˆ 0,1 𝑛 and 𝑓(π‘₯+𝑦+𝑑). Recall our target: 𝑓(π‘₯+𝑦). What’s the difference? The derivative: Ξ” 𝑑 𝑓 𝑧 =𝑓 𝑧+𝑑 𝑓(𝑧). Good: deg 2 ( Ξ” 𝑑 𝑓) ≀ deg 2 (𝑓) βˆ’1. Bad: Ξ” 𝑑 𝑓 0 ≀ 𝑓 (That’s where the factor of 2 𝑑 comes from.)

15 Goal: compute 𝑓(π‘₯+𝑦) where 𝑓: 0,1 𝑛 β†’ Β±1 πœ“ =|+βŒͺ π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“
πœ“ =|+βŒͺ π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“ Add phase πœ’ 𝛼 (𝑦) πœ“β€² Decoding + Fourier One more issue: Only Alice knows 𝑑! Bob doesn’t. It’s unaffordable to send 𝑑. Obs: 𝑠𝑒𝑝𝑝 Ξ” 𝑑 𝑓 βŠ†π΄+𝐴,βˆ€π‘‘. 𝐴=𝑠𝑒𝑝𝑝 𝑓 Thus in round 2, Alice and Bob can just encode the entire 𝐴+𝐴. 𝑑 𝑓(π‘₯+𝑦+𝑑) 𝑑 Measure{|+βŒͺ,|βˆ’βŒͺ} Measure 𝑑 A random π‘‘βˆˆ 0,1 𝑛 and 𝑓(π‘₯+𝑦+𝑑). Recall our target: 𝑓(π‘₯+𝑦). What’s the difference? The derivative: Ξ” 𝑑 𝑓 𝑧 =𝑓 𝑧+𝑑 𝑓(𝑧). Good: deg 2 ( Ξ” 𝑑 𝑓) ≀ deg 2 (𝑓) βˆ’1. Bad: Ξ” 𝑑 𝑓 0 ≀ 𝑓 (That’s where the factor of 2 𝑑 comes from.)

16 At last, deg 2 𝑓 𝑑 =0, a constant function.
Goal: compute 𝑓(π‘₯+𝑦) where 𝑓: 0,1 𝑛 β†’ Β±1 πœ“ =|+βŒͺ π›Όβˆˆπ΄ 𝑓 𝛼 πœ’ 𝛼 π‘₯ 𝐸 𝛼 πœ“ Add phase πœ’ 𝛼 (𝑦) πœ“β€² Decoding + Fourier 𝑑 𝑓(π‘₯+𝑦+𝑑) 𝑑 Measure{|+βŒͺ,|βˆ’βŒͺ} Measure 𝑑 A random π‘‘βˆˆ 0,1 𝑛 and 𝑓(π‘₯+𝑦+𝑑). Compute 𝑓 2 ≝ Ξ” 𝑑 𝑓 𝑧 =𝑓 𝑧+𝑑 𝑓(𝑧). At last, deg 2 𝑓 𝑑 =0, a constant function. Cost: log 𝐴 + log 2𝐴 + log 4𝐴 +…+ log 2 π‘‘βˆ’1 𝐴 ≀ 2 𝑑 log |𝐴| . Used trivial bound: π‘˜π΄ ≀ 𝐴 π‘˜

17 Open problems Get rid of the factor 2 𝑑 !
What can we say about additive structure of supp 𝑓 for Boolean functions 𝑓? Say, 𝐴+𝐴 ≀ 𝐴 1.5 ?

18 Thanks


Download ppt "Effcient quantum protocols for XOR functions"

Similar presentations


Ads by Google