 # A note about claw function with a small range

In the claw detection problem we are given two functions f:D→ R and g:D→ R (|D|=n, |R|=k), and we have to determine if there is exist x,y∈ D such that f(x)=g(y). We show that the quantum query complexity of this problem is between Ω(n^1/2k^1/6) and O(n^1/2+εk^1/4) when 2≤ k<n.

## Authors

##### This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

## 1 Introduction

In this note we study the problem in which given two discrete functions and (, ) we have to determine if there is a collision, i.e., inputs such that . In contrast to the problem, where the input is a single function and we have to determine if is injective, is non-trivial even when . This is the setting we focus on.

Both and have wide applications as useful subroutines in more complex algorithms [5, 12] and as a means of lower bounding complexity [10, 1].

and were first tackled by Buhrman et al. in 2000  where they gave an algorithm and lower bound. In 2003 Ambainis, introducing a novel technique of quantum walks, improved the upper bound to in the query model . It was soon realized that a similar approach works for [9, 13, 15]. Meanwhile Aaronson and Shi showed a lower bound that holds if the range . Eventually Ambainis showed that the bound holds even if . The same lower bound has since been reproved using the adversary method . Until now, only the bound based on reduction of searching was known for with .

We consider quantum query complexity of where the input functions are given as a list of their values in black box. Let denote the bounded error quantum query complexity of . For a short overview of black box model refer to Buhrman and de Wolf’s survey . Let denote . Let be defined as

Our contribution is a quantum algorithm for showing and a lower bound . In section 2 we describe the algorithm, and in section 3 we give the lower bound.

## 2 Results

###### Theorem 1.

For all , we have

###### Proof.

Let , be the inputs of the function. We denote .

Consider the following algorithm parametrized by .

1. Select a random sample of size and query the variables .
Denote by the set containing their values. Do a Grover search for an element such that . If found, output 1.

2. Select a random sample of size and query the variables .
Denote by the set containing their values. Do a Grover search for an element such that . If found, output 1.

3. Run algorithm (with the value of specified below) with the following oracle:

1. To get : do a pseudorandom permutation on using seed and using Grover’s minimum search return the first value such that .

2. To get : do a pseudorandom permutation on using seed and using Grover’s minimum search return the first value such that .

Let , be the sets containing the indices of the variables which have values not seen in the steps 1 and 1’. We denote .

Let us calculate the probability that after step 1 there exists an unseen value

which is represented in at least variables, i.e., . Consider an arbitrary value such that . For , let be the event that . . Let . Then . Using Chernoff inequality (see e.g. ),

The probability that there exists such is at most . Therefore, with probability after step , every value is represented in the input less than times. The same reasoning can be applied to step and the set . Therefore, with probability both and are at most .

Similarly, we show that with probability each appears as the first element from in at least one of the permutations of the oracle in step 2. Let be the event that appears in the -th permutation as the first element from . . Let . . . . The same argument works for . Therefore, if there is a collision, it will be found by the algorithm with probability .

We also show that with probability , in all permutations the first element from appears no further than in position (and similarly for ). We denote by the event that in the -th permutation in the -th position is an element from . . We denote . . . . Therefore, the Grover’s minimum search will use at most queries.

The steps 1 and 1’ use queries to obtain the random sample, and queries to check if there is a colliding element on the other side of the input. The oracle in step 2 uses queries to obtain one value of or .

Therefore the total complexity of the algorithm is

By using the algorithm in step 2,

 =n23β+12−12β =n12+16β =n4+ϰ−α6,

and the total complexity is minimized by setting . However, we can do better than that. Notice that the algorithm might not be the best choice for solving in step 2.

Let denote the regular algorithm. For , let denote a version of algorithm from Theorem 1 that in step 2 calls . Then we show that for all and all ,

where .

The proof is by induction on . For , we trivially have that . For the inductive step, consider the analysis of our algorithm. Let us set . First, notice that is non-decreasing in and for all . Thus for all , we have , hence and . Second, since the coefficient of is the function is above for , establishing . This confirms that is a valid choice of .

It remains to show that the complexity of step 2 does not exceed . By the inductive assumption and analysis of the algorithm, the complexity (up to logarithmic factors) of the second step is to the power of . Finally, we have to show that

By expanding and with a slight rearrangement, we obtain

We can further rearrange the required inequality by bringing to right hand side and everything else to the other. Then we get

After simplification we obtain , which is true.

Since and , the result follows. ∎

## 3 Lower Bound

We show a quantum query complexity lower bound for .

###### Theorem 2.

For all , we have .

###### Proof.

Let be the partial function defined as

Consider the function . One can straightforwardly reduce to by setting

 x′i={xi,if xi≠∗k+1,if xi=∗

and

 y′i={yi,if yi≠∗k+2,if yi=∗.

Next, we show that . The fact that has been established by Zhang . Furthermore, thanks to the work done by Brassard et al. in [6, Theorem 13] we know that for a composition theorem holds: . Therefore,

## 4 Open Problems

Can we show that ? In particular, our algorithm struggles with instances where there are singletons only two (or none) of which are matching and the remaining variables are evenly distributed with copies each, such that none are matching. Thus our algorithm then either has to waste time sampling all the high-frequency decoy values or have most variables not sampled by step 2. If this lower bound held, it would imply a better lower bound for evaluating constant depth formulas and Boolean matrix product verification [10, Theorem 5].

## References

•  Scott Aaronson, Nai-Hui Chia, Han-Hsuan Lin, Chunhao Wang, and Ruizhe Zhang. On the Quantum Complexity of Closest Pair and Related Problems. In Shubhangi Saraf, editor, 35th Computational Complexity Conference (CCC 2020), volume 169 of Leibniz International Proceedings in Informatics (LIPIcs), pages 16:1–16:43, Dagstuhl, Germany, 2020. Schloss Dagstuhl–Leibniz-Zentrum für Informatik.
•  Scott Aaronson and Yaoyun Shi. Quantum lower bounds for the collision and the element distinctness problems. Journal of the ACM (JACM), 51(4):595–605, 2004.
•  Andris Ambainis. Polynomial degree and lower bounds in quantum complexity: Collision and element distinctness with small range. Theory of Computing, 1(1):37–46, 2005.
•  Andris Ambainis. Quantum walk algorithm for element distinctness. SIAM Journal on Computing, 37(1):210–239, 2007.
•  Daniel J. Bernstein, Stacey Jeffery, Tanja Lange, and Alexander Meurer. Quantum algorithms for the subset-sum problem. In Philippe Gaborit, editor, Post-Quantum Cryptography, pages 16–33, Berlin, Heidelberg, 2013. Springer Berlin Heidelberg.
•  Gilles Brassard, Peter Høyer, Kassem Kalach, Marc Kaplan, Sophie Laplante, and Louis Salvail. Key establishment à la merkle in a quantum world. Journal of Cryptology, 32(3):601–634, July 2019.
•  Harry Buhrman and Ronald de Wolf.

Complexity measures and decision tree complexity: a survey.

Theoretical Computer Science, 288(1):21–43, 2002. Complexity and Logic.
•  Harry Buhrman, Christoph Dürr, Mark Heiligman, Peter Høyer, Frédéric Magniez, Miklos Santha, and Ronald de Wolf. Quantum algorithms for element distinctness. SIAM Journal on Computing, 34(6):1324–1330, 2005.
•  Andrew M. Childs and Jason M. Eisenberg. Quantum algorithms for subset finding. Quantum Info. Comput., 5(7):593–604, November 2005.
•  Andrew M. Childs, Shelby Kimmel, and Robin Kothari. The quantum query complexity of read-many formulas. In Proceedings of the 20th Annual European Conference on Algorithms, ESA’12, pages 337–348, Berlin, Heidelberg, 2012. Springer-Verlag.
•  Fan Chung and Linyuan Lu. Concentration inequalities and martingale inequalities: a survey. Internet Mathematics, 3(1):79–127, 2006.
•  François Le Gall and Saeed Seddighin. Quantum meets fine-grained complexity: Sublinear time quantum algorithms for string problems. 2020.
•  Frédéric Magniez, Miklos Santha, and Mario Szegedy. Quantum algorithms for the triangle problem. SIAM Journal on Computing, 37(2):413–424, 2007.
•  Ansis Rosmanis. Adversary lower bound for element distinctness with small range. 2014.
•  Seiichiro Tani. Claw finding algorithms using quantum walk. Theoretical Computer Science, 410(50):5285–5297, 2009. Mathematical Foundations of Computer Science (MFCS 2007).
•  Shengyu Zhang. Promised and distributed quantum search. In Lusheng Wang, editor, Computing and Combinatorics, pages 430–439, Berlin, Heidelberg, 2005. Springer Berlin Heidelberg.