A Glimpse of Quantum Computation

Similar documents
Quantum algorithms (CO 781, Winter 2008) Prof. Andrew Childs, University of Waterloo LECTURE 11: From random walk to quantum walk

Quantum walk algorithms

Hamiltonian simulation and solving linear systems

Quantum algorithms. Andrew Childs. Institute for Quantum Computing University of Waterloo

Discrete quantum random walks

Hamiltonian simulation with nearly optimal dependence on all parameters

Quantum algorithms (CO 781, Winter 2008) Prof. Andrew Childs, University of Waterloo LECTURE 1: Quantum circuits and the abelian QFT

Quantum NP - Cont. Classical and Quantum Computation A.Yu Kitaev, A. Shen, M. N. Vyalyi 2002

Exponential algorithmic speedup by quantum walk

arxiv: v2 [quant-ph] 7 Jan 2010

Chapter 10. Quantum algorithms

Complex numbers: a quick review. Chapter 10. Quantum algorithms. Definition: where i = 1. Polar form of z = a + b i is z = re iθ, where

Universal computation by quantum walk

Quantum algorithms (CO 781, Winter 2008) Prof. Andrew Childs, University of Waterloo LECTURE 15: Unstructured search and spatial search

arxiv:cond-mat/ v1 20 Sep 2004

High-precision quantum algorithms. Andrew Childs

Figure 1: Circuit for Simon s Algorithm. The above circuit corresponds to the following sequence of transformations.

Math 443/543 Graph Theory Notes 5: Graphs as matrices, spectral graph theory, and PageRank

Lecture 5: Random Walks and Markov Chain

Algorithmic challenges in quantum simulation. Andrew Childs University of Maryland

The Principles of Quantum Mechanics: Pt. 1

Lecture 4: Postulates of quantum mechanics

Algorithmic challenges in quantum simulation. Andrew Childs University of Maryland

Shor s Prime Factorization Algorithm

Which Quantum Walk on Graphs?

Hamiltonian simulation with nearly optimal dependence on all parameters

Quantum Computing Lecture 6. Quantum Search

Compute the Fourier transform on the first register to get x {0,1} n x 0.

By allowing randomization in the verification process, we obtain a class known as MA.

Introduction to Quantum Computing

Advanced Cryptography Quantum Algorithms Christophe Petit

Quantum expanders from any classical Cayley graph expander

Quantum algorithms based on quantum walks. Jérémie Roland (ULB - QuIC) Quantum walks Grenoble, Novembre / 39

MATH 423 Linear Algebra II Lecture 33: Diagonalization of normal operators.

The Quantum Complexity of Computing Schatten p-norms

6.842 Randomness and Computation March 3, Lecture 8

Physics 239/139 Spring 2018 Assignment 2 Solutions

C/CS/Phys 191 Quantum Gates and Universality 9/22/05 Fall 2005 Lecture 8. a b b d. w. Therefore, U preserves norms and angles (up to sign).

Quantum Computing. 6. Quantum Computer Architecture 7. Quantum Computers and Complexity

Quantum Complexity Theory and Adiabatic Computation

Quantum Computing Lecture 3. Principles of Quantum Mechanics. Anuj Dawar

Quantum algorithms based on span programs

1 Readings. 2 Unitary Operators. C/CS/Phys C191 Unitaries and Quantum Gates 9/22/09 Fall 2009 Lecture 8

Quantum algorithms (CO 781/CS 867/QIC 823, Winter 2013) Andrew Childs, University of Waterloo LECTURE 13: Query complexity and the polynomial method

PHY305: Notes on Entanglement and the Density Matrix

arxiv:quant-ph/ v2 26 Mar 2005

Quantum Algorithms for Graph Traversals and Related Problems

C/CS/Phys C191 Grover s Quantum Search Algorithm 11/06/07 Fall 2007 Lecture 21

Unitary Dynamics and Quantum Circuits

Introduction to Quantum Computing

04. Five Principles of Quantum Mechanics

QUANTUM CRYPTOGRAPHY QUANTUM COMPUTING. Philippe Grangier, Institut d'optique, Orsay. from basic principles to practical realizations.

b) (5 points) Give a simple quantum circuit that transforms the state

SVD, Power method, and Planted Graph problems (+ eigenvalues of random matrices)

Quantum Computing Lecture Notes, Extra Chapter. Hidden Subgroup Problem

Quantum Mechanics II: Examples

Errata list, Nielsen & Chuang. rrata/errata.html

1 Fundamental physical postulates. C/CS/Phys C191 Quantum Mechanics in a Nutshell I 10/04/07 Fall 2007 Lecture 12

Quantum Computing Lecture 2. Review of Linear Algebra

The Framework of Quantum Mechanics

arxiv: v1 [quant-ph] 22 Feb 2018

Factoring integers with a quantum computer

Ma/CS 6b Class 20: Spectral Graph Theory

arxiv:quant-ph/ v1 21 Nov 2003

Short Course in Quantum Information Lecture 5

Quantum Algorithms for Finding Constant-sized Sub-hypergraphs

A quantum walk based search algorithm, and its optical realisation

6.896 Quantum Complexity Theory October 2, Lecture 9

Lecture 10: Eigenvalue Estimation

Introduction to Adiabatic Quantum Computation

1 Random Walks and Electrical Networks

The query register and working memory together form the accessible memory, denoted H A. Thus the state of the algorithm is described by a vector

From Adversaries to Algorithms. Troy Lee Rutgers University

Grover s algorithm. We want to find aa. Search in an unordered database. QC oracle (as usual) Usual trick

On the query complexity of counterfeiting quantum money

1 Linear Algebra Problems

Quantum Phase Estimation using Multivalued Logic

Ph 219b/CS 219b. Exercises Due: Wednesday 4 December 2013

A Review of Quantum Random Walks and their Algorithmic Applications

1 Mathematical preliminaries

Chapter 6 Inner product spaces

Witness-preserving Amplification of QMA

Simulation of quantum computers with probabilistic models

Finite-dimensional spaces. C n is the space of n-tuples x = (x 1,..., x n ) of complex numbers. It is a Hilbert space with the inner product

Quantum Information & Quantum Computing

2. Introduction to quantum mechanics

Ph 219b/CS 219b. Exercises Due: Wednesday 22 February 2006

Lecture 2: From Classical to Quantum Model of Computation

Linear algebra and applications to graphs Part 1

α x x 0 α x x f(x) α x x α x ( 1) f(x) x f(x) x f(x) α x = α x x 2

Quantum Mechanics C (130C) Winter 2014 Final exam

Ma/CS 6b Class 20: Spectral Graph Theory

Any pure quantum state ψ (qubit) of this system can be written, up to a phase, as a superposition (linear combination of the states)

Quantum algorithm for linear systems of equations Final Report Mayank Sharma

Attempts at relativistic QM

Eigenvalues, random walks and Ramanujan graphs

Chapter 2 The Group U(1) and its Representations

Quantum Mechanics I Physics 5701

Quantum Search on Strongly Regular Graphs

PLEASE LET ME KNOW IF YOU FIND TYPOS (send to

Transcription:

A Glimpse of Quantum Computation Zhengfeng Ji (UTS:QSI) QCSS 2018, UTS 1. 1

Introduction What is quantum computation? Where does the power come from? Superposition Incompatible states can coexist Transformation of superpositions Fourier transform Quantum walk Hamiltonian simulation Intuition 2. 1

Superposition Lecture by Allan Adams https://www.youtube.com/watch?v=lz3bpuko5zc Superposition of function calls f : {0,1} n {0,1} m Assume that has a classical circuit of size polynomial in n, then the following transformation is ef cient on a quantum computer 1 2 n x x,0 1 2 n x x, f(x) 2. 2

Quantum Fourier Transform 3. 1

Let Discrete Fourier Transform (DFT) X = {0,1,,N 1} be a nite set The discrete Fourier transform transforms a function f^ f : X C to another function where f^ 1 (y) = ω xy f(x). N N x X A transformation on a sequence of N complex numbers Important tool in classical signal processing 3. 2

Fast Fourier Transform (FFT) Cooley Tukey algorithm, from f^ 1 (y) = N x X Gauss 1805! Assume for simplicity N = 2 n O( ) Recursive structure: for, N 2 to O(N log N) ω xy N 0 y < N/2 f(x). [ (y) = f^ (y)] e(y) + ω y N f^ o f^ 1 2 3. 3

What about the second half? f^ 0y ( ) 1 [ (y)] = (y) + 2 f^ e ω y N f^ o f^ 1y ( ) 1 [ (y)] = (y) 2 f^ e ω y N f^ o FFT: compute the fast Fourier transform (recursively) on the even and odd parts respectively and combine the results using the above rule 3. 4

Quantum Fourier Transform (QFT) For function f : X C, de ne quantum state f = f(x) x. x X Quantum Fourier transform maps the superposition f to f^ Recall that quantum computation is about the processing of superposition states In QFT, we transform the amplitudes by DFT The QFT is a unitary transform F = 1 N x,y X ω xy N y x 3. 5

QFT: E cient Realization Use the recursion in a quantum way f^ 0y ( ) f^ 1y ( ) 1 [ (y)] = (y) + 2 f^ e ω y N f^ o 1 [ (y)] = (y) 2 f^ e ω y N f^ o Apply QFT of qubits to the rst qubits only once!!! n 1 n 1 Then do some post-processing Cheating?! No! 3. 6

1. After the QFT on the rst qubits, we have 2. De ne function Apply n 1 f = 0 + 1 0 + 1 fe fo fe ^ fo ^ go(y) ^ = ω y (y) N f^ o n 1 controlled phase gates, we have fe ^ 0 + 1 go ^ 3. Apply Hadamard to the last qubit, we get + fe ^ go ^ 0 fe + ^ go ^ 1 2 2 4. Move the last qubit to the rst by SWAP gates and we have f^

3. 7

Remarks on QFT We have shown an ef cient implementation of QFT Complexity or Discard small rotations, gives an approximate QFT Complexity n 2 nlog n log 2 N U i V i ϵ Lemma. For unitary operators and,, we have U t U2U1 V t V2V1 tϵ. U i V i Do NOT expect it to be useful in most problems of signal processing You cannot read out the values of the amplitudes! 3. 8

Phase Estimation In phase estimation, we are given access to controlled-u x for all x X and also an eigenstate ϕ of U U ϕ iϕ = e ϕ The problem is to estimate the phase ϕ some desired precision [0, 2π) to Inverse Fourier transform F 1 1 = N xy ω N y x x,y X 3.9

n ϕ To get an -bit estimate of, the phase estimation procedure 1 x ϕ 1. Prepare the state N x X 2. Apply unitary operator x x x X U x The state becomes 1 N x X e iϕx x ϕ 3. Apply an inverse quantum Fourier transform on the rst register ϕ = 2πy/2 n y X If for some, the state in the rst register before the inverse Fourier transform is F y The rst register is a good approximation of ϕ/2π 3. 10

An Example: Discrete Log Given a cyclic group g0 generated by g0 and an element, the size is known h G The discrete logarithm of in is the smallest non- such that with respect to g0 negative integer G = N = G α log g0 h h G g α 0 = h Important for classical cryptography (Dif e- Hellman key exchange) No ef cient classical algorithm is known 3. 11

Quantum Algorithm for Discrete Log Consider the following unitary U : g hg U x It is possible to implement controlled- operators for any x by repeated squaring De ne states 1 = ϕ y ω xy, then N N gx 0 x X 1 ϕ y ω xy N g x+ log h g 0 0 ω y log h g 0 N N x X Use phase estimation! We can sample a random pair of (y, ϕ ) y Factoring is in BQP U = = ϕ y

3. 12

Quantum Walk 4. 1

Let Continuous-time Random Walk on Graph G = (V, E) be a graph The Laplacian of is a matrix of size where The continuous-time random walk on the differential equation The distribution evolves as L G V V L j,k = deg(j) 1 0 d dt p j (t) = p(t) = e Lt G k V j = k (j, k) E (j, k) E is de ned as the solution of p(0) L jk p k (t) 4. 2

Schrödinger equation: Continuous-time Quantum Walk Continuous-time quantum walk is the quantum time evolution de ned by the Laplacian From random transformations of probability distributions to transformations of superpositions Probability to amplitudes: For i d dt ψ(t) = Hψ(t) q j (t) = j ψ(t) d i (t) = (t) dt q j k V L jk q k, we have Continuous-time random and quantum walks exhibit very different properties! 4. 3

Discrete-time Random Walk In a discrete-time random walk on graph to its neighbors with equal probability G, we move from a vertex The walk is de ned by the matrix M jk ={ 1/deg(k) 0 If the probability distribution is after one step of random walk is (j, k) E otherwise p p, the distribution = Mp 4. 4

Discrete-time Quantum Walk It is not easy to come up with a proper de nition j? 1 = k j is not a unitary map! deg(j) k:(j,k) E The key idea to solves this problem is to enlarge the Hilbert space Walk on the edges, instead of the vertices (Watrous) j C = j j (2 I) j j j V j U = SC S Leaves invariant The quantum walk operator where swaps the two vertex registers 4. 5

Szegedy Walk: Markov Chain Quantized In general, for any stochastic matrix whose entry is the probability of moving from to De ne state De ne operator j=1 Quantum walk operator for one step is j N k P = j k ψ j P kj k=1 N Π = ψ j ψ j P kj U = S(2Π I) 4. 6

Spectrum Theorem Theorem. For any stochastic matrix, de ne matrix and let be the complete set of D jk = P jk P kj eigenvectors of ±i and arccos λ. D { λ } P. Then the discrete-time quantum walk corresponding to have eigenvalues U = S(2Π I) P ±1 e 4. 7

Recall that we have a Markov chain De ne isometry Check that For state k V and we de ned states λ ~, check that U = S, and US = 2λS λ ~ λ ~ λ ~ λ ~ λ ~ The two-dimensional subspace spanned by and is invariant under. Coincidence? U P = j, k ψ j P kj T = ψ j = j, k j j P kj j V j,k V T T T T = I, T = Π, ST = D = T λ λ ~ S λ ~

4. 8

Hitting Time G G M Given graph, decide if has a marked vertex in or not Consider classical Markov chain simplicity that is symmetric P P P and assume for De ne as but stops if one reaches a marked vertex P =( 0) =( ), ( P ) t P t M P M Q I Starting from a uniform distribution on not nding a marked vertex after steps is P t 0 Q I V M 1, = N M P t 1 M M PM t P M, the probability of t 4. 9

Probability of not nding a marked vertex is at most P M t A lemma that bounds the norm of P M Lemma. If the second largest eigenvalue of is at most, and, then. M ϵn 1 δϵ P M P 1 δ To bring P M t = O(1/δϵ) t to some constant below some constant, we need, so the classical hitting time is O(1/δϵ) 4. 10

Quantum Hitting Time Consider the quantum walk operator corresponding to, and State ψ 1 = ψ j N M j M The matrix for the Markov chain is U P D P ( P M0 0) U ψ Phase estimation of on state solves the decision problem If M is empty, we have P = P ψ and is a 1-eigenstate of U, and the phase estimation always outputs 0 ψ Otherwise, when M is not empty, the state lives in the ±i subspace with eigenvalues e arccos λ Time complexity: 1/arccos λ = O(1/ δϵ) steps of quantum walk I 4. 11

Quantum Search In unstructured search, one is given black-box access to a function f : S {0,1} where S is a nite set of N elements The set is the set of marked items Consider the random walk on the complete graph of vertices Spectral gap of M = {x S : f(x) = 1} Quantum hitting time P is δ = N/(N 1) Slightly different from Grover's search algorithm P N P = S S I N 1 1 N 1 O(1/ δϵ) = O( N/ M ) N 4. 12

Quantum Hamiltonian Simulation 5. 1

Schrödinger equation Hamiltonian Simulation i d dt ψ(t) = H ψ(t) H 2 n 2 n, mathematically a Hermitian matrix of size by, is called the Hamiltonian of the system and, as the equation says, it governs the dynamics of system ψ(t) = ψ(0) U t = We have for unitary operator Give Hamiltonian H and time t, nd a quantum circuit U consisting of gates such that U t poly(n, t,1/ϵ) U e iht ϵ e iht 5. 2

Product Formula If both H1 and H2 are ef ciently simulated, then so is H1 + H2 Notice that e A+B may be different from e A e B as A and B may not commute with each other Lie product formula m = O((νt e A+B = lim ( N ea/n e B/N ) N ) 2 /ϵ) ν = max(, ) For, where H1 H2, we have t/m i t/m i( + )t ( H 1 e H 1 ) m e ϵ e i H 1 H2 5. 3

Local Hamiltonians can be ef ciently simulated H = j H j term qubits H H j where each acts on at most k can be speci ed ef ciently 5. 4

An only N N Sparse Hamiltonian Simulation Hermitian matrix is sparse if in any xed row there are nonzero entries d = poly(log N) Assume there is a classical algorithm that given the row index, outputs the list of indices and values of the nonzero entries Simple case: H is diagonal (Not a small sum) x,0 x, H x,x x, e ith x,x H x,x x,0 e ith x,x = ( x ) 0 e iht 5. 5

Graph Edge Coloring Let G be an undirected graph of maximum degree An edge-coloring of at most colors can be ef ciently computed for bipartite Simulate the Hamiltonian for different color respectively and use Lie product formula y = Let be the vertex adjacent to by an edge of color De ne max x v c (x) c min = min(x, y) x = max(x, y) d 2 and V 2 2 x min, max x x H x,c Let be the by unitary acting on the space spanned by de ned by the Hamiltonian G x c d 5. 6

De ne two quantum circuits : x,0,,, x U1 min max x x V where x is a description of by unitary 2 V x V acts on the span of x and Consider U 1 and assume V x 2 :,, ϕ,, ϕ U2 min max x x V min max x x x V x V x min max x x where x 1 U2U1 = α + β V min max x x x x,0 min, max,, x x x V x min, max, x x x V x V x min max x x V min min max x x x x V max x x min max x x = α,,, + β,,, α + β By linearity U1 1 U2U1 simulate the color c part of the Hamiltonian

5. 7

More E cient Methods All known product formula based algorithms have complexity superlinear in t, is it possible to achieve linear complexity? Better performance in Connections to continuous-time and discrete-time quantum walks De ne isometry polylog 1 ϵ N j = j, k + H j, N + jk 1 ψ j X k=1 H 1 1 N jk X k=1 A unitary operator motived by the quantum walk operator that has ±i eigenvalues arccos λ e 5. 8

Phase estimation ψ 1. Apply the isometry T to state 2. Perform the phase estimation on the quantum walk operator with some precision δ 3. Compute an estimate of λ from the estimated value of 4. Introduce the phase e iλt 5. Uncompute everything Quantum signal processing Reading material: HHL algorithm Hamiltonian simulation + Phase estimation arccos λ [Low, Chuang, PRL 118 010501] [Harrow, Hassidim, Lloyd, PRL 103 150502] 5. 9

Summary A glimpse of quantum computation through Quantum Fourier transform f f^ Discrete-time quantum walk = S(2Π I) ψ t 1 ψ t ψ t 1 Hamiltonian simulation ψ(0) ψ(t) = ψ(0) e iht 6. 1