دسترسی نامحدود
برای کاربرانی که ثبت نام کرده اند
برای ارتباط با ما می توانید از طریق شماره موبایل زیر از طریق تماس و پیامک با ما در ارتباط باشید
در صورت عدم پاسخ گویی از طریق پیامک با پشتیبان در ارتباط باشید
برای کاربرانی که ثبت نام کرده اند
درصورت عدم همخوانی توضیحات با کتاب
از ساعت 7 صبح تا 10 شب
ویرایش:
نویسندگان: Helmut Bez. Tony Croft
سری: Advances in Applied Mathematics
ISBN (شابک) : 2022039303, 9781032206493
ناشر: CRC Press/Chapman & Hall
سال نشر: 2023
تعداد صفحات: 391
[392]
زبان: English
فرمت فایل : PDF (درصورت درخواست کاربر به PDF، EPUB یا AZW3 تبدیل می شود)
حجم فایل: 8 Mb
در صورت تبدیل فایل کتاب Quantum Computation به فرمت های PDF، EPUB، AZW3، MOBI و یا DJVU می توانید به پشتیبان اطلاع دهید تا فایل مورد نظر را تبدیل نمایند.
توجه داشته باشید کتاب محاسبات کوانتومی نسخه زبان اصلی می باشد و کتاب ترجمه شده به فارسی نمی باشد. وبسایت اینترنشنال لایبرری ارائه دهنده کتاب های زبان اصلی می باشد و هیچ گونه کتاب ترجمه شده یا نوشته شده به فارسی را ارائه نمی دهد.
محاسبات کوانتومی ریاضیات محاسبات کوانتومی را ارائه می دهد. هدف معرفی مبحث محاسبات کوانتومی به دانشجویان رشته های علوم کامپیوتر، فیزیک و ریاضیات است که هیچ دانش قبلی در این زمینه ندارند. کتاب در دو بخش نوشته شده است. موضوعات ریاضی اولیه مورد نیاز برای درک اولیه محاسبات کوانتومی در بخش اول مورد بررسی قرار می گیرد: مجموعه ها، توابع، اعداد مختلط و دیگر ساختارهای ریاضی مرتبط از جبر خطی و انتزاعی. موضوعات با مثالهایی با تمرکز بر جنبههای محاسباتی کوانتومی نشان داده شدهاند که در قسمت دوم با جزئیات بیشتر دنبال میشوند. بخش دوم اطلاعات کوانتومی، اندازهگیری کوانتومی و الگوریتمهای کوانتومی را مورد بحث قرار میدهد. این موضوعات پایههایی را فراهم میکنند که بر اساس آنها میتوان به موضوعات پیشرفتهتر با اطمینان نزدیک شد. امکانات رویکردی در دسترستر از اکثر متون رقیب، که برای دانشآموزان امروزی خیلی سریع به موضوعات پیشرفته و در سطح تحقیقاتی میرسند. بخش اول در ارائه تمام پشتوانه های ریاضی لازم، به ویژه برای کسانی که به فرصت بیشتری برای توسعه شایستگی ریاضی خود نیاز دارند، جامع است. دانشآموزان با اعتماد بهنفستر میتوانند مستقیماً به قسمت دوم بروند و به عنوان مرجع به قسمت اول برگردند. ایده آل برای استفاده به عنوان متن مقدماتی برای دوره های محاسبات کوانتومی. مثال های کاملاً کار شده کاربرد تکنیک های ریاضی را نشان می دهد. تمرینات در سراسر مفاهیم را توسعه می دهد و درک را افزایش می دهد. تمرینات پایان فصل تمرین بیشتری را در ایجاد یک پایه ایمن ارائه می دهد.
Quantum Computation presents the mathematics of quantum computation. The purpose is to introduce the topic of quantum computing to students in computer science, physics and mathematics who have no prior knowledge of this field. The book is written in two parts. The primary mathematical topics required for an initial understanding of quantum computation are dealt with in Part I: sets, functions, complex numbers and other relevant mathematical structures from linear and abstract algebra. Topics are illustrated with examples focussing on the quantum computational aspects which will follow in more detail in Part II. Part II discusses quantum information, quantum measurement and quantum algorithms. These topics provide foundations upon which more advanced topics may be approached with confidence. Features A more accessible approach than most competitor texts, which move into advanced, research-level topics too quickly for today's students. Part I is comprehensive in providing all necessary mathematical underpinning, particularly for those who need more opportunity to develop their mathematical competence. More confident students may move directly to Part II and dip back into Part I as a reference. Ideal for use as an introductory text for courses in quantum computing. Fully worked examples illustrate the application of mathematical techniques. Exercises throughout develop concepts and enhance understanding. End-of-chapter exercises offer more practice in developing a secure foundation.
Cover Half Title Series Page Title Page Copyright Page Dedication Contents Preface Acknowledgements Symbols I: Mathematical Foundations for Quantum Computation 1. Mathematical preliminaries 1.1. Objectives 1.2. Definitions and notation 1.3. Venn diagrams 1.4. Laws of set algebra 1.5. Boolean algebra 1.5.1. The exclusive-or operator (xor) 1.6. Groups 1.7. Cartesian product 1.8. Number bases 1.9. Modular arithmetic 1.10. Relations, equivalence relations and equivalence classes 1.10.1. Equivalence relations 1.11. Combinatorics-permutations and combinations 1.12. End-of-chapter exercises 2. Functions and their application to digital gates 2.1. Objectives 2.2. Introductory definitions and terminology 2.3. Some more functions f : R→R 2.3.1. The relative growth of functions 2.4. The Boolean functions f : B→B 2.5. Functions defined on Cartesian products 2.5.1. The Boolean functions f : Bx B→B 2.5.2. The Boolean functions f : B2→B2 2.5.3. Further Boolean functions 2.6. Further composition of functions 2.7. The Cartesian product of functions 2.8. Permuting (swapping) binary digits and binary strings 2.8.1. A classical digital circuit for swapping binary digits 2.8.2. Swapping binary digits using the Feynman cnot gate 2.8.3. Swapping strings of binary digits-vectorising the swap operator 2.9. Copying binary digits and binary strings 2.9.1. Fan-out 2.9.2. A dupe gate 2.9.3. The cnot gate 2.9.4. The Feynman double gate 2.9.5. The ccnot (Toffoli) gate 2.9.6. Fan-out copying of binary strings 2.9.7. Digital string copying with the non-reversible copy and reversible swap gates 2.9.8. Digital string copying using only reversible (the cnot and swap) gates 2.9.9. Digital string copying using only the cnot gate 2.10. Periodic functions 2.10.1. Real-valued periodic functions 2.10.2. Periodic Boolean functions 2.11. End-of-chapter exercises 3. Complex numbers 3.1. Objectives 3.2. Introduction to complex numbers and the imaginary number i 3.3. The arithmetic of complex numbers 3.4. The set of all complex numbers as a field 3.5. The Argand diagram and polar form of a complex number 3.6. The exponential form of a complex number 3.7. The Fourier transform: an application of the exponential form of a complex number 3.8. End-of-chapter exercises 4. Vectors 4.1. Objectives 4.2. Vectors: preliminary definitions 4.3. Graphical representation of two- and three-dimensional vectors 4.4. Vector arithmetic: addition, subtraction and scalar multiplication 4.5. Qubits represented as vectors 4.5.1. Combining kets and functions f : B→B 4.6. The inner product (scalar product, dot product) 4.6.1. The inner product in R2 and R3 4.6.2. The inner product on Cn: Definition 1 4.6.3. The inner product on Cn: Definition 2 4.7. End-of-chapter exercises 5. Matrices 5.1. Objectives 5.2. Matrices: preliminary definitions 5.3. Matrix arithmetic: addition, subtraction and scalar multiplication 5.4. The product of two matrices 5.5. Block multiplication of matrices 5.6. Matrices, inner products and ket notation 5.7. The determinant of a square matrix 5.8. The inverse of a square matrix 5.8.1. The inverse of an n x n matrix 5.9. Similar matrices and diagonalisation 5.10. Orthogonal matrices 5.11. Unitary matrices 5.12. Matrices and the solution of linear simultaneous equations 5.12.1. Gaussian elimination 5.13. Matrix transformations 5.14. Projections 5.15. End-of-chapter exercises 6. Vector spaces 6.1. Objectives 6.2. Definition of a vector space 6.3. Inner product spaces 6.4. Subspaces 6.5. Linear combinations, span and LinSpan 6.6. Linear independence 6.7. Basis of a vector space 6.8. Change of basis matrices 6.9. Orthogonal projections onto subspaces 6.10. Construction of an orthogonal basis - the Gram-Schmidt process 6.11. The Cartesian product of vector spaces 6.12. Equivalence classes defined on vector spaces 6.13. The sum of a vector and a subspace 6.14. The quotient space 6.15. End-of-chapter exercises 7. Eigenvalues and eigenvectors of a matrix 7.1. Objectives 7.2. Preliminary definitions 7.3. Calculation of eigenvalues 7.4. Calculation of eigenvectors 7.5. Real symmetric matrices and their eigenvalues and eigenvectors 7.6. Diagonalisation of real symmetric matrices 7.7. The spectral theorem for symmetric matrices 7.8. Self-adjoint matrices and their eigenvalues and eigenvectors 7.9. End-of-chapter exercises 8. Group theory 8.1. Objectives 8.2. Preliminary definitions and the axioms for a group 8.3. Permutation groups and symmetric groups 8.4. Unitary groups 8.5. Cosets, partitions and equivalence classes 8.6. Quotient groups 8.7. End-of-chapter exercises 9. Linear transformations 9.1. Objectives 9.2. Preliminary information 9.3. The kernel and image of a linear transformation 9.4. Linear functionals 9.5. Matrix representations of linear transformations 9.6. Bilinear maps 9.7. End-of-chapter exercises 10. Tensor product spaces 10.1. Objectives 10.2. Preliminary discussion 10.3. Calculation of tensor products 10.4. Inner products and norms on the tensor product space C2 x C2 10.5. Formal construction of the tensor product space 10.5.1. The free vector space generated by U x V 10.5.2. An equivalence relation on LinSpan(U x V) 10.5.3. Definition of the tensor product space 10.6. End-of-chapter exercises 11. Linear operators and their matrix representations 11.1. Objectives 11.2. Linear operators 11.3. The matrix representation of a linear operator 11.4. The matrix representation of a linear operator when the underlying basis is orthonormal 11.5. Eigenvalues and eigenvectors of linear operators 11.6. The adjoint and self-adjoint linear operators 11.7. Unitary operators 11.8. Linear operators on tensor product spaces 11.9. End-of-chapter exercises II: Foundations of quantum-gate computation 12. Introduction to Part II 12.1. Objectives 12.2. Computation and physics 12.3. Physical systems 12.4. An overview of digital computation 12.4.1. Digital computer states 12.4.2. Digital computer dynamics 12.4.3. Digital computer `measurement' 12.5. The emergence of quantum computation 12.6. Observables and measurement 12.6.1. Observables 12.6.2. Measurement 12.7. The Stern-Gerlach quantum experiment 13. Axioms for quantum computation 13.1. Objectives 13.2. Quantum state spaces for computation 13.2.1. The 2-dimensional vector space representation of 'spin' and quantum bits 13.2.2. The case for quantum bit (qubit) representation in C2 13.2.3. Quantum bits - or qubits 13.2.4. Global phase 13.2.5. Projective spaces and the Bloch sphere 13.2.6. Multi-qubit state spaces 13.3. Quantum observables and measurement for computation 13.4. Quantum dynamics for computation 13.5. Orthogonal projection in a complex inner product space 13.6. A summary of the axioms for quantum computation 13.7. Special cases of the measurement axiom 13.8. A formal comparison with digital computation 13.9. Gates 14. Quantum measurement 1 14.1. Objectives 14.2. Measurement using Axiom 3.2 14.2.1. Measurement of non-superimposed states in C2 14.2.2. Measurement of non-superimposed states in C2 x C2 14.2.3. Measurement of non-superimposed states in C2 x C2 x C2 15. Quantum information processing 1: the quantum emulation of familiar invertible digital gates 15.1. Objectives 15.2. On the graphical representation of quantum gates 15.3. A 1-bit/qubit gate 15.3.1. The digital not gate 15.3.2. The quantum not gate, notQ, on BC2 = {|x : x 2 B} 15.4. 2-bit/qubit gates 15.4.1. The non-invertible digital cnot, or xor, gate 15.4.2. The invertible digital cnot, or Feynman FD, gate 15.4.3. The quantum cnot, or Feynman FQ, gate on Bx2 C2 = {|x|yr : x, y 2 B} 15.5. 3-bit/qubit gates 15.5.1. The digital Toffoli (or ccnot) gate 15.5.2. The quantum Toffoli (or ccnot) gate on Bxr3C2 = {|x|y|z : x,y, z 2 B} 15.5.3. The digital Peres (invertible half-adder) gate 15.5.4. The quantum Peres gate on Bx3C2 = {|x|y|z : x, y,z B} 16. Unitary extensions of the gates notQ; FQ; TQ and PQ: more general quantum inputs 16.1. Objectives 16.2. A lemma on unitary operators 16.3. The notQ gate on C2 16.4. The Feynman FQ gate on x2C2 16.5. The quantum To oli gate on x3C2 16.6. The quantum Peres gate on x3C2 16.7. Summation expressions for the unitary extensions 16.7.1. On C2 16.7.2. On C2 x C2 16.7.3. On C2 x C2 x C2 16.7.4. Notation and closing observations 17. Quantum information processing 2: the quantum emulation of arbitrary Boolean functions 17.1. Objectives 17.2. Notation 17.3. Quantum emulation of arbitrary invertible Boolean functions 17.3.1. The quantum emulation of the invertible subset of F(B,B) 17.3.2. The quantum emulation of the invertible subset of F(B2,B2) 17.3.3. Explicit forms of the emulations of Section 17.3.2 17.3.4. The invertible subset of F(Bn,Bn); n > 3 17.4. Quantum emulation of arbitrary non-invertible Boolean functions 17.4.1. A fundamental lemma 17.4.2. The non-invertible subset of F(B,B) 17.4.3. The non-invertible functions F(B2,B) 17.4.4. The non-invertible subset of F(Bn,Bn) 17.4.5. The general functions F(Bn, Bm) 17.5. Black-box representations of Boolean functions and their quantum emulations 18. Invertible digital circuits and their quantum emulations 18.1. Objectives 18.2. Invertible digital circuits 18.3. Junk removal 18.4. Quantum emulation 19. Quantum measurement 2: general pure states, Bell states 19.1. Objectives 19.2. The measurement of super-imposed quantum states 19.3. Measuring the EPR-Bell state ½ (|0|0 + |1|1) 19.4. More general measurements of 2-qubit states in the computational basis 19.5. Measuring 2-qubit states in the Bell basis 19.5.1. Measuring the observables S2 3 and S21 20. Quantum information processing 3 20.1. Objectives 20.2. Quantum parallelism 20.3. Qubit swapping 20.4. Quantum copying - the no-cloning theorem 20.5. Quantum teleportation 1, computational-basis measurement 20.6. Quantum teleportation 2, Bell-basis measurement 21. More on quantum gates and circuits: those without digital equivalents 21.1. Objectives 21.2. General 1-qubit quantum gates 21.3. The Pauli and the not 1-qubit gates 21.4. Further 1-qubit gates: phase-shift and Hadamard 21.5. Universal 1-qubit circuits 21.6. Some 2-qubit quantum gates 21.6.1. Tensor product gates 21.6.2. The Hadamard-cnot circuit 21.7. The Hadamard gate on n-qubit registers 22. Quantum algorithms 1 22.1. Objectives 22.2. Preliminary lemmas 22.3. Diagrammatic representation of the measurement process 22.4. An introduction to computational complexity 22.5. Oracle-based complexity estimation 22.6. Deutsch's algorithm 22.7. The Deutsch-Jozsa algorithm 22.8. The Bernstein-Vazirani algorithm 23. Quantum algorithms 2: Simon's algorithm 23.1. Objectives 23.2. Periodicity, groups, subgroups and cosets 23.2.1. Real-valued functions 23.2.2. A summary of the real-valued case 23.3. Boolean functions 23.3.1. The subgroups of group (B3,) 23.3.2. The cosets B3=Ks 23.4. The hidden subgroup problem 23.5. Simon's problem 23.6. The complexity of digital solutions 23.7. Simon's quantum algorithm A. Probability A.1. Definition of probability A.2. Discrete random variables and probability distributions B. Trigonometric ratios and identities C. Coordinate systems D. Field axioms E. Solutions to selected exercises References Index