这是用户在 2024-11-29 18:13 为 https://app.immersivetranslate.com/pdf-pro/b68ead95-af8d-47fd-827e-66c7bdc29208 保存的双语快照页面，由 沉浸式翻译 提供双语支持。了解如何保存？

2024 Asia and Pacific Mathematical Contest in Modeling Wuyue Cup Quantum Computing Challenge
2024 亚太数学建模竞赛 五越杯量子计算挑战赛

Problem 问题

Exploring Frontiers in Quantum-Accelerated AI
探索量子加速 AI 的前沿

Quantum computing holds tremendous potential for solving complex problems and processing large-scale datasets, far surpassing the capabilities of classical computers. When integrated with artificial intelligence (AI), quantum computing can bring revolutionary breakthroughs. Its parallel processing capabilities enable solving more intricate problems in shorter periods, which is of significant importance for optimizing and enhancing AI algorithms. This competition aims to assess participants’ modeling and problem-solving skills in the field of quantum computing. Through challenging real-world scenarios, we explore the limitless possibilities that arise from the synergy of quantum computing and AI.
量子计算在解决复杂问题和处理大规模数据集方面具有巨大潜力，远远超过了传统计算机的能力。当与人工智能 （AI） 集成时，量子计算可以带来革命性的突破。它的并行处理能力能够在更短的时间内解决更复杂的问题，这对于优化和增强 AI 算法具有重要意义。本次竞赛旨在评估参赛者在量子计算领域的建模和解决问题的能力。通过具有挑战性的真实场景，我们探索了量子计算和 AI 协同作用带来的无限可能性。
The competition consists of two parts, each involving a dataset and a task. Participants are required to formulate a corresponding QUBO (Quadratic Unconstrained Binary Optimization) model and solve it using the simulated annealing algorithm provided by the Kaiwu SDK. The QUBO model, suitable for CPQC (Coherent Photonic Quantum Computer), is expressed as:
比赛由两部分组成，每部分都涉及一个数据集和一个任务。参赛者需要制定相应的 QUBO（Quadratic Unconstrained Binary Optimization，二次无约束二进制优化）模型，并使用 Kaiwu SDK 提供的模拟退火算法进行求解。适用于 CPQC（相干光子量子计算机）的 QUBO 模型表示为：
min x T Q x , x ∈ { 0 , 1 } n min x T Q x , x ∈ { 0 , 1 } n minx^(T)Qx,quad x in{0,1}^(n)\min x^{T} Q x, \quad x \in\{0,1\}^{n}
where Q is the coefficient matrix.
其中 Q 是系数矩阵。
The competition focuses on AI-related scenarios, converting problems into QUBO form and solving them using the Kaiwu SDK, a specialized software development kit for solving QUBO models on CPQC. The SDK is accessible at this link (https://platform.qboson.com/).
比赛侧重于 AI 相关场景，将问题转换为 QUBO 形式，并使用 Kaiwu SDK 解决这些问题，Kaiwu SDK 是一种用于在 CPQC 上解决 QUBO 模型的专用软件开发套件。可通过此链接 （https://platform.qboson.com/） 访问 SDK。
Reference materials for QUBO modeling across different scenarios are provided in the attachments to help participants understand and apply these concepts.
附件中提供了跨不同场景的 QUBO 建模的参考资料，以帮助参与者理解和应用这些概念。

Problem 1: Feature Selection for the German Credit Scoring Dataset (60 points)
问题 1：德国信用评分数据集的特征选择 （60 分）

Background 背景

Feature selection is a critical step in artificial intelligence, particularly when working with high-dimensional data. By selecting the most relevant features, we can improve the performance of predictive models while reducing complexity and overfitting. Integrating quantum computing could introduce new perspectives to feature selection problem. By converting the selection problem into QUBO form, quantum computing can accelerate the solution process.
特征选择是人工智能的关键步骤，尤其是在处理高维数据时。通过选择最相关的特征，我们可以提高预测模型的性能，同时降低复杂性和过度拟合。集成量子计算可以为特征选择问题引入新的视角。通过将选择问题转换为 QUBO 形式，量子计算可以加速求解过程。

Problem Description 问题描述

The German credit scoring dataset
德国信用评分数据集
(https://archive.ics.uci.edu/dataset/144/statlog+german+credit+data) is a classic dataset in artificial intelligence, containing features that describe various characteristics of individuals, as well as their credit status (good or bad). Your task is to perform feature selection for the German credit scoring dataset using QUBO modeling. The goal is to identify a subset of features that maximizes classification performance while minimizing the number of selected features.
（https://archive.ics.uci.edu/dataset/144/statlog+german+credit+data） 是人工智能中的经典数据集，包含描述个人各种特征及其信用状况（好或坏）的特征。您的任务是使用 QUBO 建模为德国信用评分数据集执行特征选择。目标是确定一个特征子集，该子集可最大限度地提高分类性能，同时最大限度地减少所选特征的数量。

You are expected to: 您需要：

  1. Formulate the feature selection problem described above as a QUBO model. Clearly define the objective function and decision variables.
    将上述特征选择问题表述为 QUBO 模型。明确定义目标函数和决策变量。
  2. Use Kaiwu SDK’s simulated annealing algorithm to solve the QUBO model above.
    使用 Kaiwu SDK 的模拟退火算法，解决上述 QUBO 模型。
  3. Train a classification model using the selected features and report the classification accuracy.
    使用所选特征训练分类模型并报告分类准确性。
Problem 2: Exploring the Integration of Quantum Computing and Deep Learning for Image Classification (40 points)
问题 2：探索量子计算和深度学习的集成用于图像分类 （40 分）

Background 背景

Image classification is a fundamental task in computer vision where deep learning models achieve remarkable accuracy. However, optimizing these models often requires significant computational resources. By leveraging quantum computing’s optimization capabilities, more efficient training and inference methods can be developed.
图像分类是计算机视觉中的一项基本任务，其中深度学习模型可实现极高的准确性。但是，优化这些模型通常需要大量的计算资源。通过利用量子计算的优化功能，可以开发更高效的训练和推理方法。

You are expected to: 您需要：

You are tasked with focusing specifically on image classification problem. You need to select a specific dataset and design a suitable deep learning model and structure. Then, transform the related optimization problem (e.g., model training) into a QUBO model. Solve it using Kaiwu SDK’s simulated annealing algorithm.
您的任务是专门关注图像分类问题。您需要选择特定的数据集并设计合适的深度学习模型和结构。然后，将相关的优化问题（例如模型训练）转换为 QUBO 模型。使用 Kaiwu SDK 的模拟退火算法解决。

Submission Requirements: 提交要求：

  1. Complete code files, including data preprocessing, model construction, QUBO model conversion, and solution processes. Please also include a detailed description of the code’s functionality, usage instructions.
    完整的代码文件，包括数据预处理、模型构建、QUBO 模型转换和求解过程。请还包括代码功能的详细说明、使用说明。
  2. A detailed result report containing:
    详细的结果报告，其中包含：
    a. The problem scenario and its background.
    一个。问题场景及其背景。
    b. Description of the artificial intelligence model structure.
    b.人工智能模型结构的描述。
    c. Specific QUBO model formulation.
    c. 特定的 QUBO 模型制定。
    d. Solution process and solving time of the QUBO model, together with results analysis.
    d. QUBO 模型的求解过程和求解时间，以及结果分析。
    f. Model performance evaluation and interpretation of results.
    f.模型性能评估和结果解释。
Please provide sources and explanations for any external datasets or reference materials used.
请为所使用的任何外部数据集或参考资料提供来源和解释。
Note: 注意：
  1. Kaiwu SDK is restricted to solve problem under 600 bits.
    开物 SDK 仅限于解决 600 位以下的问题。
  2. Only the latest version of Kaiwu SDK is supported.
    仅支持最新版本的 Kaiwu SDK。
  3. For the Problem 1, participants may choose to submit their matrices to CPQC (https://platform.qboson.com/), which is optional. Each participant will receive a total of 5 quotas during the competition. No additional quotas will be granted once they are used up. Additionally, there may be delays in receiving results from CPQC.
    对于问题 1，参与者可以选择将他们的矩阵提交给 CPQC （https://platform.qboson.com/），这是可选的。每位参赛者在比赛期间将获得总共 5 个名额。配额用完后将不再授予。此外，从 CPQC 接收结果可能会有延迟。
  4. If you have any questions about the tasks, please scan the QR code below to contact us.
    如果您对任务有任何疑问，请扫描下方二维码与我们联系。
  5. You can view the Kaiwu SDK installation and usage guide via this link (https://b23.tv/IqKoPnv).
    您可以通过此链接 （https://b23.tv/IqKoPnv） 查看 Kaiwu SDK 安装和使用指南。