QSBench/QSBench-Thermal-Demo-v1.0.0
收藏资源简介：
--- license: cc-by-nc-4.0 task_categories: - tabular-regression - feature-extraction language: - en tags: - qiskit - quantum-circuits - synthetic-dataset - benchmark - expectation-values - quantum-computing - qml-benchmark - quantum dataset - qml dataset - quantum benchmark - noisy quantum data - thermal relaxation - T1 T2 noise - error mitigation - noise robustness pretty_name: QSBench Thermal Relaxation Demo v1.0.0 – Noisy Quantum Dataset (T1/T2 Thermal Noise, n=8) size_categories: - 1K<n<10K ---  🌐 [Website](https://qsbench.github.io) | 🤗 [Dataset](https://huggingface.co/datasets/QSBench/QSBench-Thermal-Demo-v1.0.0) | 🛠️ [GitHub](https://github.com/QSBench/QSBench-Thermal-Demo-v1.0.0) | 🚀 [Interactive Demo](https://huggingface.co/QSBench/spaces) # QSBench Thermal Relaxation Demo v1.0.0 **Quantum Machine Learning dataset for realistic noise robustness and sim-to-real research.** Includes paired ideal and noisy expectation values under **thermal relaxation** (T1/T2) noise — the most physically relevant noise model on current quantum hardware. **2048 high-quality synthetic quantum circuits with thermal relaxation noise** — demo subset of the QSBench Noise Pack. Designed for researchers and engineers working on noise-aware quantum ML, decoherence analysis, and error mitigation. ### Why this dataset? Real quantum hardware is dominated by **energy relaxation (T1)** and **dephasing (T2)**. Thermal relaxation is the standard model used by IBM, Google, IonQ and most superconducting platforms. This dataset allows you to: - Study realistic **T1/T2 decoherence** effects - Train models that predict or mitigate relaxation-induced errors - Benchmark **robustness** under physically accurate noise - Explore sim-to-real transfer on 8-qubit circuits ### Use Cases - Noise robustness benchmarking (T1/T2) - Error mitigation research (especially relaxation errors) - Predicting noisy expectation values - Learning error correction models - Feature engineering for decoherence-aware quantum ML ### Dataset Overview - **Samples**: 2048 - **Qubits**: 8 - **Depth**: 6 - **Circuit Families**: Mixed (HEA, RealAmplitudes, QFT, Efficient SU(2), Random) - **Entanglement**: Full - **Noise**: Thermal Relaxation (`T1 = 50 μs`, `T2 = 30 μs`) - **Observables**: Z, X, Y in mixed mode (global + per-qubit) - **Shots**: 1024 - **Splits**: Train / Validation / Test — deterministic hash-based ### What's Inside Each Sample Each sample in the Parquet files contains: - Raw and transpiled QASM representations - Circuit adjacency matrix - Gate statistics (CX, H, RX, RY, RZ, etc.) - Structural metrics: Gate entropy + Meyer-Wallach entanglement - **Ideal expectation values** - **Noisy expectation values** (after thermal relaxation) - **Explicit error targets**: `error_<label> = ideal - noisy` - Circuit metadata and generation parameters - Deterministic split label ### Key Learning Signals For every observable, the dataset provides: `ideal_expval_*`, `noisy_expval_*`, `error_*`, `sign_ideal_*`, `sign_noisy_*`. This supports both high-precision regression and classification tasks for realistic noise modeling. ### QSBench-Thermal: Realistic Decoherence Modeling **You don't need a PhD in Quantum Physics to use this dataset.** This dataset represents the dominant noise channel on today's quantum processors — thermal relaxation. It behaves like gradual "battery drain" of quantum information over time. ### The ML Mission: Complex Tabular Regression Unlike symmetric depolarizing noise, thermal relaxation is **biased** and time-dependent. Your mission is to build a predictive model that understands how circuit structure (adjacency, gate counts, entanglement) influences T1/T2 decay rates. ### Dataset Anatomy (Features & Targets) | Group | Column Name | What is it for ML? | |----------------|------------------------------|--------------------| | **Features (X)** | `adjacency` | Graph structure — dense graphs lose coherence faster | | **Features (X)** | `qasm_transpiled` | Hardware-specific compiled circuit (NLP feature) | | **Features (X)** | `single_qubit_gates`, `two_qubit_gates` | Operation counts | | **Target (y)** | `error_Z_global`, `error_X_global` | Continuous regression targets (relaxation loss) | | **Physics** | `meyer_wallach` | Entanglement measure — does high entanglement decay faster? | ### Quick Start Idea Compare feature importances. Does `meyer_wallach` or `two_qubit_gates` matter more when predicting thermal relaxation errors? ### Load the Dataset ```python from datasets import load_dataset # Load the thermal relaxation demo dataset dataset = load_dataset("QSBench/QSBench-Thermal-Demo-v1.0.0", split="train") # Inspect the first sample print(dataset[0]) ``` ### Repository Structure The dataset is stored in the `main` branch and contains only the data files: ```text QSBench-Thermal-Demo-v1.0.0/ ├── README.md └── data/ └── shards/ └── *.parquet └── *.csv ``` All metadata files (`meta.json`, `schema.json`, `coverage.json`, etc.) are in the `metadata` branch. 👉 browse [metadata branch](https://huggingface.co/datasets/QSBench/QSBench-Thermal-Demo-v1.0.0/tree/metadata) ### Related QSBench Datasets - [QSBench-Core-v1.0.0-demo](https://huggingface.co/datasets/QSBench/QSBench-Core-v1.0.0-demo) - [QSBench-Depolarizing-Demo-v1.0.0](https://huggingface.co/datasets/QSBench/QSBench-Depolarizing-Demo-v1.0.0) - [QSBench-Amplitude-v1.0.0-demo](https://huggingface.co/datasets/QSBench/QSBench-Amplitude-v1.0.0-demo) - [QSBench-Transpilation-v1.0.0-demo](https://huggingface.co/datasets/QSBench/QSBench-Transpilation-v1.0.0-demo) ### Part of the QSBench Family This is a small public **demo version**. Full-scale datasets (20k–150k+ samples), specialized noisy versions, and custom hardware packs are available on the [QSBench website](https://qsbench.github.io/). ### Notes - Fully synthetic dataset generated with Qiskit Aer - No real-world or personal data - **License**: CC BY-NC 4.0 (Personal & Research Use) **Questions or custom requests?** Visit [QSBench website](https://qsbench.github.io/) or open an issue on [GitHub](https://github.com/QSBench/QSBench-Thermal-Demo-v1.0.0). ### Support QSBench You can support the project directly on this Giveth page: **[https://giveth.io/project/qsbench](https://giveth.io/project/qsbench)** Your donations help us generate larger datasets, cover GPU costs, and continue developing new realistic noise models. --- *Generated with QSBench Generator v5.1.0*
--- 许可证：CC BY-NC 4.0（知识共享署名-非商业性使用4.0国际协议） 任务类别： - 表格回归（tabular-regression） - 特征提取（feature-extraction） 语言：英语 标签： - Qiskit（qiskit） - 量子电路（quantum-circuits） - 合成数据集（synthetic-dataset） - 基准测试（benchmark） - 期望值（expectation-values） - 量子计算（quantum-computing） - 量子机器学习基准测试（qml-benchmark） - 量子数据集（quantum dataset） - 量子机器学习数据集（qml dataset） - 量子基准测试（quantum benchmark） - 带噪量子数据（noisy quantum data） - 热弛豫（thermal relaxation） - T1/T2噪声（T1 T2 noise） - 误差缓解（error mitigation） - 噪声鲁棒性（noise robustness） 数据集名称：QSBench Thermal Relaxation Demo v1.0.0 – Noisy Quantum Dataset (T1/T2 Thermal Noise, n=8) 样本量范围：1K<n<10K ---  🌐 [官网](https://qsbench.github.io) | 🤗 [数据集页面](https://huggingface.co/datasets/QSBench/QSBench-Thermal-Demo-v1.0.0) | 🛠️ [GitHub仓库](https://github.com/QSBench/QSBench-Thermal-Demo-v1.0.0) | 🚀 [交互式演示](https://huggingface.co/QSBench/spaces) # QSBench 热弛豫演示版v1.0.0 **面向真实噪声鲁棒性研究与仿真-实物迁移（sim-to-real）的量子机器学习数据集** 包含**热弛豫（thermal relaxation，T1/T2噪声）**下的成对理想期望值与带噪期望值——这是当前量子硬件上最具物理相关性的噪声模型。 包含2048条高质量带热弛豫噪声的合成量子电路——为QSBench噪声套件的演示子集。 本数据集专为关注噪声感知量子机器学习、退相干分析与误差缓解的研究人员与工程师打造。 ### 为何选择本数据集？ 真实量子硬件的主要噪声源为**能量弛豫（T1）**与**退相（T2）**。热弛豫是IBM、Google、IonQ以及多数超导量子平台所采用的标准噪声模型。 借助本数据集，你可以： - 研究真实的**T1/T2退相干（T1/T2 decoherence）**效应 - 训练可预测或缓解弛豫诱导误差的模型 - 在符合物理真实性的噪声环境下对**鲁棒性（robustness）**进行基准测试 - 在8量子比特电路上探索仿真-实物迁移（sim-to-real transfer） ### 应用场景 - 噪声鲁棒性基准测试（T1/T2噪声场景） - 误差缓解研究（尤其针对弛豫误差） - 带噪期望值预测 - 误差校正模型学习 - 面向退相干感知量子机器学习的特征工程 ### 数据集概览 - **样本量**：2048 - **量子比特数**：8 - **电路深度**：6 - **电路家族**：混合类型（HEA、RealAmplitudes、量子傅里叶变换QFT、高效SU(2)、随机电路） - **纠缠方式**：全纠缠 - **噪声类型**：热弛豫（`T1 = 50 μs`，`T2 = 30 μs`） - **可观测量**：Z、X、Y混合模式（全局+单比特） - **采样次数**：1024 - **数据集划分**：训练集/验证集/测试集——基于哈希的确定性划分 ### 单样本内容说明 Parquet格式文件中的每个样本包含以下内容： - 原始与编译后的量子汇编语言（QASM）表示 - 电路邻接矩阵 - 量子门统计信息（CX、H、RX、RY、RZ等） - 结构指标：门熵与Meyer-Wallach纠缠度 - **理想期望值** - **带噪期望值**（经热弛豫噪声扰动后） - **显式误差目标**：`error_<标签> = 理想值 - 带噪值` - 电路元数据与生成参数 - 确定性划分标签 ### 核心学习信号 针对每一种可观测量，本数据集均提供以下字段：`ideal_expval_*`、`noisy_expval_*`、`error_*`、`sign_ideal_*`、`sign_noisy_*`。该设计可支持面向真实噪声建模的高精度回归与分类任务。 ### QSBench-热弛豫：真实退相干建模 **无需量子物理博士学位即可使用本数据集。** 本数据集还原了当前量子处理器的主导噪声通道——热弛豫，其效应类似于量子信息随时间逐渐“电量耗尽”的过程。 ### 机器学习任务：复杂表格回归 与对称去极化噪声不同，热弛豫噪声具备**偏置性**且随时间变化。你的任务是构建一个预测模型，以理解电路结构（邻接关系、门数量、纠缠程度）如何影响T1/T2衰减速率。 ### 数据集结构（特征与目标变量） | 分组 | 字段名 | 机器学习用途 | |----------------|------------------------------|--------------------| | **特征集（X）** | `adjacency` | 电路图结构——稠密图的相干性衰减更快 | | **特征集（X）** | `qasm_transpiled` | 硬件适配的编译后电路（类自然语言特征） | | **特征集（X）** | `single_qubit_gates`, `two_qubit_gates` | 单量子比特门、双量子比特门的操作计数 | | **目标变量（y）** | `error_Z_global`, `error_X_global` | 连续回归目标（弛豫误差） | | **物理指标** | `meyer_wallach` | 纠缠度度量——高纠缠度的电路是否衰减更快？ | ### 快速探索思路 对比特征重要性：在预测热弛豫误差时，`meyer_wallach`与`two_qubit_gates`哪个的影响更显著？ ### 数据集加载示例 python from datasets import load_dataset # 加载热弛豫演示数据集 dataset = load_dataset("QSBench/QSBench-Thermal-Demo-v1.0.0", split="train") # 查看第一个样本 print(dataset[0]) ### 仓库结构 数据集存储于`main`分支，仅包含数据文件： text QSBench-Thermal-Demo-v1.0.0/ ├── README.md └── data/ └── shards/ └── *.parquet └── *.csv 所有元数据文件（`meta.json`、`schema.json`、`coverage.json`等）均存储于`metadata`分支。 👉 浏览[元数据分支](https://huggingface.co/datasets/QSBench/QSBench-Thermal-Demo-v1.0.0/tree/metadata) ### 相关QSBench数据集 - [QSBench-Core-v1.0.0-demo](https://huggingface.co/datasets/QSBench/QSBench-Core-v1.0.0-demo) - [QSBench-Depolarizing-Demo-v1.0.0](https://huggingface.co/datasets/QSBench/QSBench-Depolarizing-Demo-v1.0.0) - [QSBench-Amplitude-v1.0.0-demo](https://huggingface.co/datasets/QSBench/QSBench-Amplitude-v1.0.0-demo) - [QSBench-Transpilation-v1.0.0-demo](https://huggingface.co/datasets/QSBench/QSBench-Transpilation-v1.0.0-demo) ### 属于QSBench系列数据集 本数据集为公开的**精简演示版本**。全量数据集（20000–150000+样本）、专用带噪版本与定制硬件套件可在[QSBench官网](https://qsbench.github.io/)获取。 ### 注意事项 - 本数据集为完全合成生成，基于Qiskit Aer生成 - 未包含真实世界数据或个人数据 - **许可证**：CC BY-NC 4.0（仅可用于个人与研究用途） **如有疑问或定制需求**，请访问[QSBench官网](https://qsbench.github.io/)或在[GitHub仓库](https://github.com/QSBench/QSBench-Thermal-Demo-v1.0.0)提交Issue。 ### 支持QSBench项目 可通过以下Giveth页面直接支持本项目： **[https://giveth.io/project/qsbench](https://giveth.io/project/qsbench)** 您的捐赠将用于支持更大规模数据集的生成、GPU成本覆盖以及新型真实噪声模型的持续研发。 --- *由QSBench生成器v5.1.0生成*
