教育>本科研究生>计算机类
自动机器学习  

自动机器学习  "

作者:张峰
ISBN:9787121457050
定价:¥79.0
字数:301千字
页数:188
出版时间:2023-06
开本:16开
版次:01-01
装帧:
出版社:电子工业出版社
简介

本书重点讲解基于云平台的超参数优化、神经构架搜索以及算法选择等内容,是自动机器学习的基本任务。介绍了基于三个主要云服务提供商(包括 Microsoft Azure、Amazon Web Services (AWS) 和 Google Cloud Platform)进行 AutoML,同时部署 ML 模型和管道,具有较强的实用性。在应用场景中评估 AutoML 方面,例如算法选择、自动特征化和超参数调整,并区分云和 OSS 产品等。本书适用于从事机器学习或人工智能方向的数据科学家或工程师学习,也适合学生或行业初学者进行入门学习实践。

前言

前 言 每个机器学习工程师都会和具有超参数的系统打交道,自动机器学习中最基本的任务就是自动设置这些超参数以优化性能。最新的深度神经网络在其架构、正则化和优化方面有大量的超参数,自动机器学习可以有效地进行超参数设定以节省时间和精力。 本书介绍了自动特征工程、模型和超参数调优、基于梯度的方法等基础技术,同时,带领读者探索在开源工具中实现这些技术的不同方式。此外,本书还从企业工具出发,了解在三个主要云服务提供商中使用自动机器学习的不同方法:Microsoft Azure、Amazon Web Services(AWS)和 Google Cloud Platform(GCP)。随着学习的深入,进一步通过使用自动机器学习构建机器学习模型来探索云自动机器学习平台的功能。后面的章节将展示如何通过自动机器学习来开发耗时长、包含重复任务的精准模型。 通过学习,读者能够准确构建和部署机器学习模型,还能提高生产力,实现互操作性并最大限度减少特征工程任务的自动机器学习模型。 对于数据科学家、机器学习开发人员、人工智能爱好者,或任何希望使用开源工具、Microsoft Azure 机器学习、AWS 和谷歌云平台提供的功能自动构建机器学习模型的读者,本书都可提供一定的学习指导。 本书内容 第 1 章,走进自动机器学习 从机器学习开发生命周期开始讲解,并介绍自动机器学习解决的超参数优化问题。为入门者提供自动机器学习的概述,也可为经验丰富的机器学习从业者提供参考。 第 2 章,自动机器学习、算法和技术 讲解数据科学家在没有丰富经验的情况下如何构建 AI 解决方案。分别从三方面回顾了自动机器学习当前的发展:自动特征工程、超参数优化及神经架构搜索。并介绍了在这三个类别中所采用的最先进技术,包括贝叶斯优化、强化学习、进化算法和基于梯度的方法。最后总结流行的自动机器学习框架并总结自动机器学习当前面临的挑战。 第 3 章,使用开源工具和库进行自动机器学习 介绍自动机器学习开源软件(OSS)工具和库,这些工具和库可自动化预测模型的构思、概念化、开发和部署的整个生命周期。从数据准备到模型训练再到验证和部署,这些工具几乎可以在无人工干预的情况下完成所有工作。同时,回顾主要的 OSS 工具,包括TPOT、AutoKeras、auto-sklearn、Featuretools、Microsoft NNI 和 AutoGluon,并帮助读者了解每个库中所采用的不同价值主张和方法。 第 4 章,Azure Machine Learning 本章介绍 Azure 机器学习,利用 Windows Azure 平台和服务的强大功能可加速端到端机器学习生命周期。在本章中,我们将回顾如何开始使用企业级机器学习服务来构建和部署模型,从而使开发人员和数据科学家能够更快地构建、训练和部署机器学习模型。通过示例,我们将为构建和部署自动机器学习解决方案奠定基础。 第 5 章,使用 Azure 进行自动机器学习 通过代码示例详细介绍了如何使用 Azure 机器学习栈自动执行模型开发中耗时长的迭代任务,并使用 Azure AutoML 执行回归、分类和时间序列分析等操作,执行超参数调优以找到最佳参数,并使用 Azure AutoML 找到最佳模型。 第 6 章,使用 AWS 进行机器学习 介绍了 Amazon SageMaker Studio、Amazon SageMaker Autopilot、Amazon SageMaker Ground Truth 和 Amazon SageMaker Neo,以及 AWS 提供的其他 AI 服务和框架。除了超大规模调参(云产品),AWS 还提供广泛和深入的机器学习服务和云基础设施,使得开发人员、数据科学家和资深从业者都能方便使用机器学习。AWS 提供机器学习服务、人工智能服务、深度学习框架和学习工具来快速构建、训练和部署机器学习模型。 第 7 章,使用 Amazon SageMaker Autopilot 进行自动机器学习 进一步介绍 Amazon SageMaker Studio,使用 SageMaker Autopilot 运行多个候选对象,以找出数据预处理步骤、机器学习算法和超参数的最佳组合。此外,提供了训练一个推理流水作业(inference pipeline)的动手、说明性概述,以便在实时端点或批处理上轻松部署。 第 8 章,使用 Google Cloud Platform 进行机器学习 介绍了 Google 的 AI 和机器学习产品。Google Cloud 在可信且可扩展的平台上提供创新的机器学习产品和服务。这些服务包括 AI Hub、AI 构建块(如视觉、语言、对话和结构化数据服务)及 AI Platform。通过这一章的学习,读者可以熟悉这些产品并了解 AI Platform如何支持 Google 的开源平台 Kubeflow,该平台使开发人员能够构建便携式机器学习流水作业(pipeline),并可以访问尖端的 Google AI 技术,例如,TensorFlow、TPU 和 TFX 工具,将 AI 应用程序部署到生产环境中。 第 9 章,使用 GCP 进行自动机器学习 本章展示了如何以最少的工作量和机器学习专业知识训练自定义的特定于业务的机器学习模型。通过动手示例和代码演练,探索 Google Cloud AutoML 平台,以在自然语言、视觉、非结构化数据、语言翻译和视频智能方面创建定制的深度学习模型,而无须任何数据科学或编程知识。 第 10 章,企业中的自动机器学习 将企业环境中的自动机器学习作为一个系统,通过生成包含数据分析、预测模型和性能对比的全自动报告来实现数据科学的自动化。自动机器学习的一个独特之处在于它提供了对结果的自然语言描述,适用于机器学习领域的非专家。强调了 MLOps 管道的可操作性,并讨论在实际问题上表现良好的方法,并确定最佳整体方案。介绍了现实世界中挑战背后的想法和概念,并提供了解决这些问题的思路。 学习要求 本书是对自动机器学习的介绍。熟悉数据科学、机器学习和深度学习方法将有助于了解自动机器学习如何改进现有方法。

目录

第 1 章 走进自动机器学习··············································································.1 1.1 机器学习开发生命周期 ·······································································.1 1.2 自动机器学习简介 ·············································································.2 1.3 自动机器学习的工作原理 ····································································.3 1.4 数据科学的大众化 ·············································································.5 1.5 揭穿自动机器学习的迷思 ····································································.5 1.6 自动机器学习生态系统 ·······································································.6 1.7 小结 ·······························································································11 第 2 章 自动机器学习、算法和技术··································································12 2.1 自动机器学习概述 ·············································································12 2.2 自动特征工程 ···················································································15 2.3 超参数优化 ······················································································16 2.4 神经架构搜索 ···················································································18 2.5 小结 ·······························································································19 第 3 章 使用开源工具和库进行自动机器学习······················································20 3.1 技术要求 ·························································································20 3.2 自动机器学习的开源生态系统 ······························································21 3.3 TPOT······························································································22 3.4 Featuretools ······················································································29 3.5 Microsoft NNI ···················································································32 3.6 auto-sklearn ······················································································38 3.7 AutoKeras ························································································41 3.8 Ludwig ····························································································44 3.9 AutoGluon························································································44 3.10 小结······························································································44 第 4 章 Azure Machine Learning········································································45 4.1 Azure Machine Learning 入门 ································································45 4.2 Azure Machine Learning 栈 ···································································46 4.3 Azure Machine Learning 服务 ································································50 4.4 使用 Azure Machine Learning 建模 ·························································56 4.5 使用 Azure Machine Learning 部署和测试模型 ··········································68 4.6 小结 ·······························································································70 第 5 章 使用 Azure 进行自动机器学习 ·······························································71 5.1 Azure 中的自动机器学习 ·····································································71 5.2 使用自动机器学习进行时间序列预测 ·····················································85 5.3 小结 ·······························································································97 第 6 章 使用 AWS 进行机器学习 ······································································98 6.1 AWS 环境中的机器学习······································································98 6.2 开始使用 AWS ···············································································.101 6.3 使用 Amazon SageMaker Autopilot·······················································.109 6.4 使用 Amazon SageMaker JumpStart······················································.111 6.5 小结 ····························································································.111 第 7 章 使用 Amazon SageMaker Autopilot 进行自动机器学习······························.113 7.1 技术要求 ······················································································.113 7.2 创建 Amazon SageMaker Autopilot 受限实验··········································.113 7.3 创建 AutoML 实验 ··········································································.120 7.4 运行 SageMaker Autopilot 实验并部署模型············································.123 7.5 构建并运行 SageMaker Autopilot 实验··················································.126 7.6 小结 ····························································································.132 第 8 章 使用 Google Cloud Platform 进行机器学习 ·············································.134 8.1 Google Cloud Platform 使用入门 ·························································.134 8.2 使用 GCP 实现 AI 和 ML··································································.137 8.3 Google Cloud AI Platform 和 AI Hub·····················································.139 8.4 Google Cloud AI Platform 使用入门 ·····················································.141 8.5 使用 Google Cloud 进行 AutoML ························································.144 8.6 小结 ····························································································.146 第 9 章 使用 GCP 进行自动机器学习······························································.147 9.1 Google Cloud AutoML Tables······························································.147 9.2 创建 AutoML Tables 实验··································································.148 9.3 了解 AutoML Tables 模型部署····························································.158 9.4 在 AutoML Tables 上使用 BigQuery 公共数据集 ·····································.162 9.5 自动机器学习做价格预测 ·································································.164 9.6 小结 ····························································································.170 第 10 章 企业中的自动机器学习 ···································································.171 10.1 企业是否需要自动机器学习·····························································.171 10.2 自动机器学习——企业高级分析的加速器···········································.172 10.3 自动机器学习的挑战和机遇·····························································.173 10.4 建立信任——自动机器学习中的模型可解释性和透明度 ························.174 10.5 在企业中引入自动机器学习·····························································.176 10.6 总结与展望··················································································.177

作者简介

编辑推荐

作者寄语

电子资料

www.luweidong.cn

下一个