Programming Machine Learning Coding-From Deep Learning

《Programming Machine Learning》是一本专著旨在为开发者提供机器学习的基础知识。它不仅帮助读者理解这一领域的基本原理，并且使他们能够开发自己的机器学习系统。书中采用Python语言作为编程工具，并从线性回归入手逐步深入讲解复杂的技术细节。通过这些内容的学习与实践操作, 读者将掌握监督学习、神经网络以及深度学习等核心方法, 并学会编写完整的算法框架以解决实际问题。此外, 所有代码均基于基础编写, 以便全面解析每一条代码的功能与作用

目录

Early Praise for Programming Machine Learning

• 引言部分，包括多个对本书的高度评价。

Contents

• 致谢
  • How Odd It Seems. (机器学习的魅力何在.)

Part I — From Zero to Image Recognition

How Machine Learning Works (机器学习的工作原理)

• 程序编程与机器学习对比研究
• 监督学习方法论
• 后台数学模型解析
• 系统架构搭建指南

Your First Learning Program (你的第一个学习程序)

• 深入理解问题
  • 开发线性回归模型
  • 引入偏差项

Walking the Gradient (梯度下降)

• 我们的算法未能达到预期效果
• 梯度下降法
• 动手实践：基准站点超前操作

4. 实践：基准站点超前操作

Hyperspace! (高维空间)

• Introducing Additional Dimensions
  • Matrix Operations
  • Enhancing the Learner’s Proficiency
  • Engaging in Field-Level Statistical Roles

A Discerning Machine (辨别机器)

Despite its widespread application, linear regression faces challenges in accurately modeling complex relationships between variables. The introduction of sigmoid functions revolutionizes the field by introducing smooth, non-linear transformations that better capture intricate patterns in data. Practical applications highlight the importance of classification techniques in distinguishing between distinct categories based on input features. 6. 实践应用中的关键点

Getting Real (走向现实)

• Data is placed in the leading position (数据处于首位的位置)
  • Our dedicated MNIST repository (我们的专用MNIST数据库)
  • The actual matter is vital for practical implementation (实际的应用对于实践操作来说至关重要)

The Final Challenge (最后的挑战)

 * Going Multiclass (多类分类)
 * Moment of Truth (真相时刻)

The Perceptron (感知器)

• Introduce the concept of the Perceptron (介绍感知器)
• Construct a network using perceptrons (构建使用感知器的网络)
• Examine the limitations and shortcomings of perceptrons (分析感知器的局限性与不足之处)

Part II — Neural Networks

Designing the Network (设计网络)

 * Assembling a Neural Network from Perceptrons (用感知器组装神经网络)
 * Enter the Softmax (引入Softmax)
 * Hands On: Network Adventures (实践：网络冒险)

Building the Network (构建网络)

• Code 前向传播
• Cross entropy

• Hands-on practice: Time travel testing

Training the Network (训练网络)

• Exploring the Case for Backpropagation: A Comprehensive Analysis (深入探讨反向传播的重要性)
  • Implementing and Applying Backpropagation: Enhancing Model Training Efficiency (优化模型训练效率的方法与实践)

How Classifiers Work (分类器的工作原理)

 * Tracing a Boundary (追踪边界)
 * Bending the Boundary (弯曲边界)

Batchin’ Up (批量处理)

 * Learning, Visualized (可视化学习)
 * Batch by Batch (逐批处理)

The Zen of Testing (测试的禅意)

 * The Threat of Overfitting (过拟合的威胁)
 * A Testing Conundrum (测试难题)

Let’s Do Development (开始开发)

 * Preparing Data (准备数据)
 * Tuning Hyperparameters (调整超参数)

Part III — Deep Learning

A Deeper Kind of Network (更深层的网络)

Building a Neural Network with Keras (利用Keras构建神经网络)

Defeating Overfitting (战胜过拟合)

 * Overfitting Explained (解释过拟合)
 * Regularizing the Model (正则化模型)

Taming Deep Networks (驯服深度网络)

• Grasping the Concept of Activation Functions (掌握激活函数的概念)
• Incorporating Advanced Techniques into Your Arsenal (融入你的武库之中)

Beyond Vanilla Networks (超越基本网络)

• 《CIFAR-10数据集》（CIFAR-10数据集的来源）
• 卷积神经网络的核心模块（The Building Blocks of CNNs的同义表达）

Into the Deep (深入探索)

• Deep Learning's Emergence (深度学习的兴起)
• Extraordinary Effectiveness (非凡的效果)

Appendices

A1. Essential Python (核心Python) - 展示Python的独特之处 (展示Python的独特之处) - 深入探索其核心组件 (深入探索其核心组件)

A2. The Words of Machine Learning (机器学习的术语) - Index (索引)