Variable time amplitude amplification and quantum algorithms for linear algebra. Andris Ambainis University of Latvia

Similar documents
Principle Component Analysis

Definition of Tracking

GAUSS ELIMINATION. Consider the following system of algebraic linear equations

Remember: Project Proposals are due April 11.

DCDM BUSINESS SCHOOL NUMERICAL METHODS (COS 233-8) Solutions to Assignment 3. x f(x)

Dennis Bricker, 2001 Dept of Industrial Engineering The University of Iowa. MDP: Taxi page 1

Chapter Newton-Raphson Method of Solving a Nonlinear Equation

Partially Observable Systems. 1 Partially Observable Markov Decision Process (POMDP) Formalism

The Schur-Cohn Algorithm

Course Review Introduction to Computer Methods

Least squares. Václav Hlaváč. Czech Technical University in Prague

UNIVERSITY OF IOANNINA DEPARTMENT OF ECONOMICS. M.Sc. in Economics MICROECONOMIC THEORY I. Problem Set II

Chapter Newton-Raphson Method of Solving a Nonlinear Equation

Applied Statistics Qualifier Examination

Model Fitting and Robust Regression Methods

Rank One Update And the Google Matrix by Al Bernstein Signal Science, LLC

Multiple view geometry

Generalized Spectral Resolution & some of its applications

Dynamic Power Management in a Mobile Multimedia System with Guaranteed Quality-of-Service

Haddow s Experiment:

Proof that if Voting is Perfect in One Dimension, then the First. Eigenvector Extracted from the Double-Centered Transformed

In this Chapter. Chap. 3 Markov chains and hidden Markov models. Probabilistic Models. Example: CpG Islands

ψ ij has the eigenvalue

Chapter 1: Fundamentals

Jens Siebel (University of Applied Sciences Kaiserslautern) An Interactive Introduction to Complex Numbers

Review of linear algebra. Nuno Vasconcelos UCSD

Quadratic speedup for unstructured search - Grover s Al-

Lecture 4: Piecewise Cubic Interpolation

v v at 1 2 d vit at v v 2a d

Section 6.4 Graphs of the sine and cosine functions

The Trapezoidal Rule

CISE 301: Numerical Methods Lecture 5, Topic 4 Least Squares, Curve Fitting

Strong Gravity and the BKL Conjecture

Reproducing Kernel Hilbert Space for. Penalized Regression Multi-Predictors: Case in Longitudinal Data

Lesson 1: Quadratic Equations

Introduction to Numerical Integration Part II

CS 188: Artificial Intelligence Spring 2007

First Midterm Examination

6 Roots of Equations: Open Methods

M/G/1/GD/ / System. ! Pollaczek-Khinchin (PK) Equation. ! Steady-state probabilities. ! Finding L, W q, W. ! π 0 = 1 ρ

4. Eccentric axial loading, cross-section core

Abhilasha Classes Class- XII Date: SOLUTION (Chap - 9,10,12) MM 50 Mob no

Chapter 6 Continuous Random Variables and Distributions

Before we can begin Ch. 3 on Radicals, we need to be familiar with perfect squares, cubes, etc. Try and do as many as you can without a calculator!!!

Sections 1.3, 7.1, and 9.2: Properties of Exponents and Radical Notation

Grover s Algorithm + Quantum Zeno Effect + Vaidman

ORDINARY DIFFERENTIAL EQUATIONS

The graphs of Rational Functions

We will see what is meant by standard form very shortly

Support vector machines for regression

Statistics 423 Midterm Examination Winter 2009

Cramer-Rao Lower Bound for a Nonlinear Filtering Problem with Multiplicative Measurement Errors and Forcing Noise

Research Article On the Upper Bounds of Eigenvalues for a Class of Systems of Ordinary Differential Equations with Higher Order

QUADRATIC EQUATIONS OBJECTIVE PROBLEMS

DATA Search I 魏忠钰. 复旦大学大数据学院 School of Data Science, Fudan University. March 7 th, 2018

Two Coefficients of the Dyson Product

First Midterm Examination

p-adic Egyptian Fractions

MTH 122 Fall 2008 Essex County College Division of Mathematics Handout Version 10 1 October 14, 2008

1. Extend QR downwards to meet the x-axis at U(6, 0). y

SOLVING SYSTEMS OF EQUATIONS, ITERATIVE METHODS

Surface maps into free groups

A-Level Mathematics Transition Task (compulsory for all maths students and all further maths student)

Read section 3.3, 3.4 Announcements:

LCM AND HCF. Type - I. Type - III. Type - II

Constructing Free Energy Approximations and GBP Algorithms

Riemann is the Mann! (But Lebesgue may besgue to differ.)

ARITHMETIC OPERATIONS. The real numbers have the following properties: a b c ab ac

4.1. Probability Density Functions

Math 113 Exam 2 Practice

4 7x =250; 5 3x =500; Read section 3.3, 3.4 Announcements: Bell Ringer: Use your calculator to solve

Chapter 5 1. = on [ 1, 2] 1. Let gx ( ) e x. . The derivative of g is g ( x) e 1

Acceptance Sampling by Attributes

SUMMER KNOWHOW STUDY AND LEARNING CENTRE

Homogeneous and Heterogeneous Traffic of Data Packets on Complex Networks: The Traffic Congestion Phenomenon *

Advanced Algorithmic Problem Solving Le 3 Arithmetic. Fredrik Heintz Dept of Computer and Information Science Linköping University

APPROXIMATE INTEGRATION

The 7 th Balkan Conference on Operational Research BACOR 05 Constanta, May 2005, Romania STEADY-STATE SOLUTIONS OF MARKOV CHAINS

8. INVERSE Z-TRANSFORM

ME 501A Seminar in Engineering Analysis Page 1

5.7 Improper Integrals

Name Solutions to Test 3 November 8, 2017

Chapter Gauss-Seidel Method

Learning Enhancement Team

Numerical Integration

CHI-SQUARE DIVERGENCE AND MINIMIZATION PROBLEM

4.6 Numerical Integration

Geometric Sequences. Geometric Sequence a sequence whose consecutive terms have a common ratio.

FUNDAMENTALS ON ALGEBRA MATRICES AND DETERMINANTS

MA 15910, Lessons 2a and 2b Introduction to Functions Algebra: Sections 3.5 and 7.4 Calculus: Sections 1.2 and 2.1

3/6/00. Reading Assignments. Outline. Hidden Markov Models: Explanation and Model Learning

INTERPOLATION(1) ELM1222 Numerical Analysis. ELM1222 Numerical Analysis Dr Muharrem Mercimek

Fall 2012 Analysis of Experimental Measurements B. Eisenstein/rev. S. Errede

5.2 Exponent Properties Involving Quotients

The area under the graph of f and above the x-axis between a and b is denoted by. f(x) dx. π O

Problem Set 9. Figure 1: Diagram. This picture is a rough sketch of the 4 parabolas that give us the area that we need to find. The equations are:

Quantum Mechanics Qualifying Exam - August 2016 Notes and Instructions

AQA Further Pure 2. Hyperbolic Functions. Section 2: The inverse hyperbolic functions

Special Numbers, Factors and Multiples

Substitution Matrices and Alignment Statistics. Substitution Matrices

Transcription:

Vrble tme mpltude mplfcton nd quntum lgorthms for lner lgebr Andrs Ambns Unversty of Ltv

Tlk outlne. ew verson of mpltude mplfcton;. Quntum lgorthm for testng f A s sngulr; 3. Quntum lgorthm for solvng A=b (s n HHL08).

Vrble tme mpltude mplfcton

Ampltude mplfcton [Brssrd, Hoyer, Mosc, Tpp, 00] Algorthm A tht succeeds wth probblty >0. Success s verfble. Incresng success probblty to 3/4: Clssclly: O(/). Quntumly: O(/ ).

Serch [Grover, 96]????? Fnd n object wth certn property. Check rndom object: success probblty /. Success probblty 3/4: O(/ )=O( ) repettons.

Vrble tme quntum lgorthms Algorthm tht stops t one of severl tmes T,, T k, wth probbltes p,, p k. A A A 3 A k T T T k

Ampltude mplfcton A U A- V A- V A Runnng tme:

Our result Let T v k p T Quntum lgorthm wth success probblty nd verge runnng tme T v quntum lgorthm wth success probblty /3 nd runnng tme

Bsc de A 3 outcomes: success, flure, dd not stop Amplfy success nd dd not stop. Amplfed verson A.

Bsc de () A A 3 outcomes: success, flure, dd not stop Amplfy success nd dd not stop. Amplfed verson A.

Dffcultes Ampltude mplfcton repeted k tmes; If one mplfcton loses fctor of c, then k mplfctons lose fctor of c k. We need very precse nlyss of mpltude mplfcton.

Testng f mtr s sngulr

Sngulrty testng Mtr A; Promse A s sngulr or ll sngulr vlues of A re t lest mn. Tsk: dstngush between the two cses.

The nturl lgorthm Replce A by A 0 If A hs sngulr vlue, B hs egenvlues. Implement B s Hmltonn. B Use egenvlue estmton for B. 0 A

Egenvlue estmton Input: Hmltonn B, stte : B =. Output: estmte for. Assume tht. To obtn estmte wth -, t suffces to use B for tme O(/).

Egenvlue estmton Input: Hmltonn B, rndom stte. Output: estmte for rndom, wth -.

The nturl lgorthm () B s ether sngulr or hs mn for ll egenvlues. To test B for sngulrty: Choose = mn /3; Egenvlue estmton on rndom. If B sngulr, mn /3, wth probblty /. If B nonsngulr, mn /3. Amplfy mn /3.

Runnng tme O(/ mn ) for egenvlue estmton; Amplfcton: repettons; Totl O(/ mn ). Cn we do better f most egenvlues re substntlly more thn mn?

Our mprovement Let,,, be the egenvlues of B. Theorem There s quntum lgorthm for sngulrty wth runnng tme where ~ O v v m, mn

The lgorthm Generte rndom stte. Run egenvlue estmton severl tmes, wth precson = /3, /6,, mn /3. If estmte stsfes >, stop, output 0. If ll estmtes stsfy, output =0.

Runnng tme If >0, the lgorthm stops fter frst egenvlue estmton wth < /. O(/) steps. If =0, the egenvlue estmton s run untl = mn /3. O(/ mn ) steps. Averge runnng tme: m mn,

Solvng systems of lner equtons

The problem Gven j nd b, fnd. Best clsscl lgorthm: O(.37 ). b b b

Obstcles to quntum lgorthm b b b Obstcle : tkes tme O( ) to red ll j. Soluton: query ccess to j. Grover: serch tems wth O() quntum queres. Obstcle : tkes tme O() to output ll.

Hrrow, Hssdm, Lloyd, 008 Output = b b b Mesurement wth probblty. Estmtng c +c ++c. Seems to be dffcult clssclly.

Hrrow, Hssdm, Lloyd, 008 Runnng tme for producng : O(log c ), but wth dependence on two other prmeters. b b b

Runnng tme. Sze of system O(log c ).. Tme to mplement A O() for sprse mtrces A, O() generlly. 3. Condton number of A. k m mn m nd mn bggest nd smllest sngulr vlues of A Tme O c log

Emple A = b, A rndom mtr; Lrgest sngulr vlue: O( ); Smllest sngulr vlue: Ω(/ ). Condton number: O(). Runnng tme of HHL: O( log c ) = O( log c ).

Emple A = b, A Lplcn of d-dmensonl grd; d, j j jd djcent A j j j d d j j j d, d d 0 otherwse Condton number: O( /d ). HHL runnng tme: O( log c ) = O( 4/d log c ).

Our result Theorem There s quntum lgorthm for genertng n tme O(k +o() log c ). [HHL, 008]: (k -o() ) tme requred, unless BQP=PSPACE.

The mn des Esy-to-prepre Soluton b b b

The mn des A b b b b A b b A b

The mn des A b b We desgn physcl system wth Hmltonn A. Untry e A. e A A - v egenvlue estmton.

The mn des Let v nd be the egenvectors nd egenvlues of A. c v b c v Implement quntum trnsformton v v b

Egenvlue estmton v v ' v ' v ' ' EE EE - ot untry. v ' fl C succ C v ' ' ' Soluton: perform Amplfy.

Our lgorthm Observton: suffces to hve such tht -. Run egenvlue estmton severl tmes, wth precson = /3, /6,, mn /3. Stop when <. Generte C C. Amplfy. succ ' ' fl

Open problem Wht problems cn we reduce to systems of lner equtons (wth s the nswer)? Emples: Serch; Perfect mtchngs n grph; Grph bprtteness. Bggest ssue: condton number.