![]() | |
|
I am a professor emeritus of CS at Kanagawa Institute of Technology, Japan. Originally my specialty was parallel and distributed systems. My current interests include machine learning, natural language processing, creating mobile apps with MIT App Inventor, and quantum computing. In the web version of this blog, clicking the icon on the right (a plastic sphere) will take you to the "List of Quantum Computing Articles". - Fujio Yamamoto (for e-mail, add "@ieee.org" after "yamamotof")
2025年5月9日金曜日
最新版Qiskit2.0を利用した量子コンピューティングの解説書
2025年5月5日月曜日
IBM Quantum無料枠の利用可能時間の計算が変更になった!
IBM Quantum無料利用枠の新しい計算方法:最近気付いたことだが、私にとってはちょっと重要事項なので書き留める!
![]() |
Here’s the image illustrating the change in IBM Quantum’s free usage calculation method. |
![]() |
An example |
2025年4月29日火曜日
御礼:ブログページビュー(アクセス数)20万回超え
🔴感謝:本ブログへのアクセス回数が20万回に到達
2016年末に開設したこのブログ、このほど、20万回webビュー(アクセス)に達した。直近の約130件(これまでの総計は約500件)はほとんど量子コンピューティングに関する記事である。ブログアイコンもそれらしくした。ご愛読、そして、フィードバックしていただいた方々に御礼申し上げたい。
・掲載サイト:Linkedin記事
2025年4月28日月曜日
A Basic Mathematics Book for Learning Quantum Computing
In the previous article, we introduced a new book on quantum computing. However, many people may want to first learn basic mathematics before reading such books. I would like to briefly introduce the following book as one such book. Generally speaking, the relationship between this book and the previous one is as follows:
This book is a large volume of 539 pages. It provides a very thorough explanation of the basics of mathematics related to quantum computing. Parts 1 and 2 are basic mathematics, mainly linear algebra. However, since Dirac Notation (bra-ket) is already used here, it becomes clear that this is not purely basic mathematics, but is aimed at quantum computing. The authors explain that readers who have already mastered the basics of linear algebra can skip these parts and move on to Part 3. Even for such readers, Part 2 is very useful for reviewing points that they may have forgotten. In other words, this book also serves as an encyclopedia.
In the third and fourth parts, the most important operations in quantum computing are explained in detail, with a focus on "Tensor products". Although it is not very noticeable, it is worth noting that the "Change of Basis" introduced in the second part is explained in more detail in this third part. This will be important in many fields, including quantum key distribution later. You will also see that the "Kronecker Product" is important in simplifying quantum computing. More advanced content such as "Singular Value Decomposition" is also included. Furthermore, one of the outstanding features of this book is that "Probability", another foundation of quantum computing, is dealt with extensively in the fourth part.
At the beginning, there is a "Level Indicator" explanation, which is useful for understanding the level of difficulty of the content. However, it would be even better if it had a marking to indicate which of the minimum necessary knowladge is required to read the second book, "Quantum Computing & Information." This is because this book contains a huge amount of content, and some people want to study efficiently. For example, it may be okay to skip "Discrete Fourier Transform" and "Markov Chains" for the time being.
2025年4月24日木曜日
Visualizing and Understanding QSVC
- Normalization (scaling) of input data: No need for visualization.
- Reduction of input data dimension: Not necessary in this case, since it is two-dimensional (2 features).
- Display of input data: Draw data samples in a 2D scatter plot. → Figure 1
- Convert input data to quantum state: Display quantum state on Bloch sphere. → Figure 2
- Create quantum kernel matrix: Represented as a heat map. → Figure 3
- Perform SVC learning: Display the resulting classification decision boundary in 2D. → Figure 4
2025年4月23日水曜日
量子サポートベクタ分類QSVCのビジュアル化
- 入力データの正規化(スケーリング):ビジュアル化の必要は特になし。
- 入力データの次元削減:今回は2次元(2 features)なので不要。
- 入力データの表示:データサンプルを2D散布図に描く。→図2
- 入力データを量子状態に変換:ブロッホ球に量子状態を表示。→図3
- 量子カーネル行列の作成:ヒートマップとして表現。→図4
- SVC学習を行う:結果としての分類決定境界を2D表示。→図5
2025年4月18日金曜日
量子機械学習の第一歩QSVC(Quantum Support Vector Classifier)
【要旨】古典的手法では見つけられなかった(or できなかった)ことが、量子的手法では可能になったという事例があれば勇気付けられる。その一つとして、量子計算を用いたQSVC(Quantum Support Vector Classifier)が、古典SVCよりも高い精度で分類できる一つの例を示す。この例に対しては、古典SVCでは見つからなかった、新たな分類決定境界が、量子QSVCで見つけられたと言える。これは、量子機械学習に取り組む上で意義がある事例と思われる。
🔴対象としたデータセットGaussian-Parity
今回用いたデータセットGaussian-Parityの一例を図1に示す。右側の表は、その内容である。2つの特徴量(Feature1、Feature2)とラベル(0 or 1)で構成される300サンプルが含まれる。これをMiniMax(0〜π)スケーリングして散布図にしたものが左側の図である。ラベル0とラベル1のデータが、斜めに交差して分布しているので、境界線を引くのは難しそうに見える。
🔴古典的SVCによるクラス分け
このデータセットに対して、まず、古典SVCでクラス分けした。300サンプルの7割を訓練用として学習させた結果を、残り3割のサンプルをテスト用として評価した分類の精度は81%であった。学習結果を反映した分類の決定境界を図2に示した。かなりよく分類できていると思われる。SVCの威力が感じられた。(SVCに与える種々のパラメータの値で結果は変動するが。)
次に、量子的QSVCでクラス分けした。今回は、Qiskitに装備されているライブラリQSVCを量子回路シミュレーションで実行した。学習の条件はSVCの場合と同じである。分類精度は、90%となり、SVCの場合よりもかなり高まった。それは、図3に示す通り、図2とは異なる決定境界が得られたことによる。このように、SVCでは見つからなかった新たな決定境界が得られた理由は、(技術詳細は略すが)少ない量子ビットnでも、2のn乗次元の広い量子状態の空間を探索できることによるのであろう。すなわち、多様なカーネル行列と呼ばれる情報を古典的なSVCの仕組みに与えることができる。
🔴古典的SVCと量子的QSVCとの関係
詳細は、別のブログ記事で後日議論したいが、図4に両者の関係の概要を示した。Javier Mancilla M.氏の著書"QML Unlocked"が非常に参考になったので、その中のFigure 16を引用し、加筆した。(この書籍のレビュー結果の記事も別途書く予定である。)
2025年4月9日水曜日
Reduction of error occurrence in IBM quantum computer (Heron processor)
Theoretically, the success rate of Mermin-Peres magic, a 4-qubit application using quantum entanglement, is 100%. When this was executed on ibm_torino (Heron r1), the success rate was 92% due to errors caused by noise. This is a big improvement from the 86% success rate on ibm_brisbane (Eagle r3) a while ago. The reason for this is that the error rate was 14% on ibm_brisbane, but was reduced to 7.9% on ibm_torino. Figure 1 shows the details.
2025年4月8日火曜日
IBM量子コンピュータ(Heronプロセッサ)におけるエラー発生の低減
【要旨】量子もつれを用いた例題Mermin-Peres-Magicを、IBM量子コンピュータ新鋭機ibm_torino (Heron r1プロセッサ)で実行させた結果、ノイズ等によるエラーの発生率が、以前のibm_brisbane (Eagle r3プロセッサ)に比べて、ほぼ半減することが分かった。これにより、所望の正解が得られる成功率は、86%から92%に大きく改善された。今後出現するであろう華々しい成果は、このような長年の地道な研究開発によるものなのだと実感できた気がする。
🔴誤り低減を目指す量子コンピュータの進展
Eagle(鷲)は攻撃的で強く、Heron(鷺)はしなやかで強いというイメージがあるという。IBMがそれを念頭において量子コンピュータに命名したのかは定かではない。IBMはこれまで、無償で量子プロセッサEagle r3(マシン名 ibm_kyiv, ibm_brisbane)を提供してきたが、この3月から、新鋭機Heron r1(マシン名 ibm_torino)を追加した。Heronは、Eaglelよりも、大幅にエラー発生率が低減されて強力になったという。そのハードウェアの仕組みは私には分からないが、最大の難題の一つであるエラー低減に向けて着実に進展していることが窺える。
2025年3月26日水曜日
Introduction to two books on quantum computing
The number of books on quantum computing has been steadily increasing. Here, I’d like to briefly introduce two books I recently purchased, both of which I found to be excellent.
🔴Peter Y. Lee, Huiwen Ji, Ran Cheng: Quantum Computing and Information, Polaris QCI Pub., 2nd edition, Feb. 2025
First off, I was excited to see the publication date listed as 2025—and the book certainly lives up to that excitement! With 502 pages, its thickness alone hints at the density of its content. The material ranges from beginner to intermediate levels and is presented with numerous illustrations, making it a great choice for those who want to study carefully and in depth.
![]() |
Published in February 2025, 502 pages total, chapter titles provided by me |
![]() |
Qiskit quantum app worked perfectly on my local environment |
🔴Jack D. Hidary: Quantum Computing - An Applied Approach, 2nd edition, Springer, Aug. 2021
This second book is also a thick, full-color hardcover with a total of 422 pages, targeting beginners to intermediate learners. The latter 170 pages provide a detailed introduction to the basics under the section titled Mathematical Tools for Quantum Computing, so it may be a good idea to review this part first.
The quantum algorithm section in the first half overlaps somewhat with the content of the aforementioned book by Lee et al., but I was happy to see that it includes QFT (Quantum Fourier Transform) and Shor’s Algorithm, which were not covered in that book. For instance, it walks through the entire process (quantum + classical parts) of factoring the small integer 15 (= 3×5), which greatly enhances reader's understanding of the Shor algorithm. Moreover, the second edition includes new content related to Quantum Machine Learning, which is a big plus.
(Note) Factoring a slightly larger integer like 184573 (= 487×379) is considered difficult in simulations due to memory limitations, and on current quantum computers due to limitations in qubit count and error rates. However, simulation is still possible by replacing "quantum order finding" with "classical order finding."
![]() |
Published in August 2021, 422 pages total |
By the way, when you open this book, the Schrödinger equation appears right at the beginning—in other words, the wave function and the Hamiltonian. Since most books on quantum computing are computer science–oriented, this equation is rarely mentioned, which is a bit surprising. However, this is actually a very good thing, because both the quantum gate model and the quantum annealing method are said to originate from the time-dependent Schrödinger equation. Though the explanation spans only four pages, it clearly demonstrates, with a few equations, that “to understand the time evolution of a wave function, one must consider the total energy of the system.” Specifically, it derives the case of a harmonic oscillator potential, as shown below.
![]() |
Schrödinger Equation – Harmonic Oscillator Potential |
Also, please note that the quantum algorithm programs provided in the book are written in Google’s Cirq, not IBM’s Qiskit as mentioned earlier. These programs were created in an older Cirq environment than mine, and I encountered various errors when trying to run them. However, the official Google Cirq site provides examples very similar to those in the book, and I was able to run them perfectly in my Cirq setup! Using those as a reference, you should be able to run the remaining examples in the book as needed.
![]() |
Examples from Google Cirq |
🔴Note the differences between IBM and Google machines
Lastly, there's something important to keep in mind: the bit order in quantum registers is reversed in IBM Qiskit (Little Endian) and Google Cirq (Big Endian)! This is an easy mistake to make and can cause all sorts of confusion. For example, if the initial state of three qubits is |000⟩, applying X gates to q1 and q2 results in q0q1q2 → 011 in Cirq, but q2q1q0 → 110 in Qiskit. For more details, please refer to my earlier article, Testing My Mobile Quantum Circuit Simulator.
tinyurl------- https://tinyurl.com/eaacxan2 -------