The first translated book on Python neural network programming is finally here

Click to follow asynchronous books , stick to the public number

Share IT good books, technical dry goods and workplace knowledge with you every day

Participate in the topic discussion at the end of the article, and give asynchronous books every day

- Asynchronous editor

Yes, the first translation book of Python neural network programming is finally here, it is "Python Neural Network Programming" , based on Python 3.5, the software tools used in the book are free and open source, no need to pay any fees, no You need an expensive computer to make your own neural network . The first time I saw this book was at the key product defense meeting 3 months ago, and there is no doubt that I passed it with a high score. This book ranks second only to the AI bible "Deep Learning" in the machine learning category in Meiya , and readers commented "a good introductory book on machine learning" and "an excellent interpretation of neural networks".

Neural networks are a topic that both artificial intelligence technology and deep learning cannot avoid.

There's a dearth of great work in making complex topics clear and accessible, allowing beginners to deepen their understanding of the technology. The book "Python Neural Network Programming" satisfies exactly these two aspects, making complex topics clear and easy to understand, and using a simple Python neural network programming example to clarify the topic.

What is a neural network? What is the use?

Neural network is a kind of neural network that simulates the human brain in order to realize artificial intelligence-like machine learning technology.

Neural network is an important machine learning technology, and it is the basis of deep learning, the hottest research direction at present.

Learning about neural networks will not only give you a powerful approach to machine learning, but it will also help you better understand deep learning techniques.

What is this book about?

The main purpose of this book is to expose the concepts behind neural networks to as many people as possible. There are 3 main parts:

The first part, an overview of the mathematical ideas used in simple neural networks;

The second part, learn Python, and learn how to use Python 3.5 to implement your own neural network, recognize handwritten digits, and test the performance of the neural network. ;

The third part, further understand the simple neural network model, observe the inside of the trained neural network, try to further improve the performance of the neural network, and deepen the understanding of related knowledge. .

This book is suitable for readers who want to engage in neural network research and exploration, as well as readers who are interested in related fields such as artificial intelligence, machine learning, and deep learning.

Misunderstanding of most people

For many years, ordinary people have had a misunderstanding about artificial intelligence, that is, artificial intelligence is just using higher-level and more complex mathematical instructions to tell computers what to do, how to simulate human behavior, and let the computer "pretend" to understand human emotions. However, the author of this book tells us that, in fact, it is better to teach a "computer" to fish than to teach a "computer" to fish. Without too advanced mathematical thinking, we can create an expert-level "neural network" with only high school mathematics . This is not an exaggeration or alarmism, but a real, real fact.

Now, major newspapers, websites, and all kinds of self-media are proclaiming a point of view, which is to warn young people to study hard, otherwise it will be difficult to find a job in the future. I think this view is too optimistic, which misleads readers into thinking that as long as they study hard now, they can successfully "counterattack". If this problem is described in a bit brain-burning and pedantic language, in a word, it is "the age of artificial intelligence has a problem of changing the way human values are reflected" . In other words, if we still rely on the knowledge in textbooks for survival, not for innovation, and not for exploration, then no matter how well we master the knowledge, we are only picking up wisdom and can only be lost under the rolling wheels of history. If you want to know why I sigh so, please read this book carefully. As long as you have a little basic mathematics in middle school, can understand Chinese, and have a little interest in computing, you can read this book. The basis of logic is actually quite simple.

content

Chapter 1 How Neural Networks Work 001

1.1 feet are shorter, inches are longer001

1.2 A simple prediction machine 003

1.3 Classifiers are not much different from predictors 008

1.4 Training a Simple Classifier 011

1.5 Sometimes a classifier is not enough to solve the problem 020

1.6 Neurons - Nature's Computing Machine 024

1.7 Tracking Signals in Neural Networks 033

1.8 By heart, matrix multiplication is useful 037

1.9 An example of a three-layer neural network using matrix multiplication 043

1.10 Learning weights from multiple nodes 051

1.11 Backpropagation Errors from Multiple Output Nodes 053

1.12 Backpropagating Errors into More Layers 054

1.13 Backpropagating Errors Using Matrix Multiplication 058

1.14 How we actually update the weights 061

1.15 Successful weight update example 077

1.16 Preparing data 078

Chapter 2 DIY with Python 083

2.1　Python　083

2.2 Interactive Python = IPython 084

2.3 Getting started with Python gracefully 085

2.4 Making Neural Networks in Python 105

2.5 Handwritten Digits Dataset MNIST 121

Chapter 3 Interesting 153

3.1 Your own handwritten numbers 153

3.2 Neural Networks Inside the Brain 156

3.3 Create new training data: rotate the image 160

3.4 Conclusion 164

Appendix A Introduction to Calculus 165

A.1 A straight line 166

A.2 A slash 168

A.3 A Curve 170

A.4 Hand Drawn Calculus 172

A.5 Non-Hand-drawn Calculus 174

A.6 Calculus without Graphs 177

A.7 Mode 180

A.8 Functions of Functions 182

Appendix B Working with a Raspberry Pi 186

B.1 Install IPython 187

B.2 Ensuring that all work is carried out normally 193

B.3 Training and Testing Neural Networks 194

B.4 Raspberry Pi succeeded 195

Using Python Neural Networks to Recognize Human Handwritten Characters

For thousands of years, humans have tried to understand the mechanics of intelligence and replicate it on thinking machines. And never content with having mechanical or electronic devices help with simple tasks, such as starting a fire with flint, lifting heavy rocks with pulleys, or doing arithmetic with a calculator.

Instead, we hope to automate more challenging, relatively complex tasks like grouping similar photos, identifying diseased cells from healthy cells, or even playing an elegant game of chess. These tasks seem to require human intelligence, or at least some deeper, more arcane ability in the human mind that cannot be found in a machine as simple as a calculator.

具有类似人类智能的机器是一个如此诱人且强大的想法，我们的文化对它充满了幻想和恐惧，如斯坦利·库布里克导演的《2001: A Space Odyssey》中的HAL 9000（拥有巨大的能力却最终给人类带来了威胁）、动作片中疯狂的“终结者（Terminator）”机器人以及电视剧《Knight Rider》中具有冷静个性的话匣子KITT汽车。

1997年，国际象棋卫冕世界冠军、国际象棋特级大师加里·卡斯帕罗夫被IBM“深蓝”计算机击败，我们在庆祝这一历史性成就的同时，也担心机器智能的潜力。

我们如此渴望智能机器，以至于一些人受到了诱惑，使用欺骗手段，例如，臭名昭著的国际象棋机器Turkey仅仅是使用一个人隐藏在机柜内而已！

在20世纪50年代，人工智能这门学科正式成立，此时，人类雄心勃勃，对人工智能抱着非常乐观的态度。最初的成功，让人们看到了计算机可以进行简单的博弈、证明定理，因此，一些人相信，在十年左右的时间内，人类级别的人工智能将会出现。

但是，实践证明：发展人工智能困难重重，进展一度停滞不前。20世纪70年代，人们在学术界挑战人工智能的雄心遭到了毁灭性的打击。接下来，人们削减了人工智能研究经费，对人工智能的兴趣消失殆尽。机器那冰冷的逻辑，绝对的1和0，看起来似乎永远不能够实现细致入微的、有机的，有时甚至模糊的生物大脑思维过程。

在一段时间内，人类未能独具匠心，百尺竿头，更进一步，将机器智能探索带出其既定轨迹。在此之后，研究人员灵光一现，尝试通过复制生物大脑工作的机制，来构建人工大脑？真正的大脑具有神经元，而不是逻辑门。真正人脑具有更优雅更有机的推理，而不是冰冷的、非黑即白的、绝对的传统算法。

蜜蜂或鸽子大脑的简单性与其能够执行复杂任务的巨大反差，这一点启发了科学家。就是这零点几克的大脑，看起来就能够做许多事情，如导航、适应风向、识别食物和捕食者、快速地决定是战斗还是逃跑。当今的计算机拥有大量的廉价资源，能够模仿和改进这些大脑吗？一只蜜蜂大约有950 000个神经元，今天的计算机，具有G比特和T比特的资源，能够表现得比蜜蜂更优秀吗？

但是，如果使用传统的方法来求解问题，那么即使计算机拥有巨大的存储和超快的处理器，也无法实现鸟和蜜蜂使用相对微小的大脑所做的事情。受到仿生智能计算的驱动，神经网络（Netural Network）出现了，并且神经网络从此成为在人工智能领域中最强大、最有用的方法之一。

今天，谷歌的Deepmind以神经网络为基础，能够做一些非常奇妙的事情，如让计算机学习如何玩视频游戏，并且在人类历史上第一次在极其变化多端的围棋博弈中击败了世界级的大师。如今，神经网络已经成为了日常技术的核心，例如自动车牌号码识别、解码手写的邮政编码。

《Python神经网络编程》所探讨的就是神经网络，让你了解神经网络如何工作，帮你制作出自己的神经网络，训练神经网络来识别人类的手写字符。如果使用传统的方法来执行这个任务，那么将是非常困难的。

《Python神经网络编程》

[英] 塔里克·拉希德（Tariq Rashid）著

点击封面购买纸书

当前，深度学习和人工智能的发展和应用给人们留下了深刻的印象。神经网络是深度学习和人工智能的关键元素，然而，真正了解神经网络工作机制的人少之又少。本书用轻松的笔触，一步一步揭示了神经网络的数学思想，并介绍如何使用Python编程语言开发神经网络。本书将带领您进行一场妙趣横生却又有条不紊的旅行——从一个非常简单的想法开始，逐步理解神经网络的工作机制。您无需任何超出中学范围的数学知识，并且本书还给出易于理解的微积分简介。

本书为美亚五星畅销书，备受关注。基于Python3.5，全彩印刷，如果只选一本神经网络图书，他是首选。

本书的目标是让尽可能多的普通读者理解神经网络。读者将学习使用Python开发自己的神经网络，训练它识别手写数字，甚至可以与专业的神经网络相媲美。本书适合想要了解深度学习、人工智能和神经网络的读者阅读，尤其适合想要通过Python编程进行神经网络开发的读者参考。

《文本上的算法——深入浅出自然语言处理》

路彦雄著

点击封面购买纸书

微信整合搜索算法组组长路彦雄全新作品，深入浅出讲解自然语言处理和机器学习技术，微博总阅读量超30万次。

本书结合作者多年学习和从事自然语言处理相关工作的经验，力图用生动形象的方式深入浅出地介绍自然语言处理的理论、方法和技术。本书抛弃掉繁琐的证明，提取出算法的核心，本书前面章节介绍了学习机器学习需要掌握的一些数学基础，帮助读者尽快地掌握自然语言处理所必备的知识和技能。本书适合从事自然语言处理相关研究和工作的读者参考，尤其适合想要了解和掌握机器学习或者自然语言处理技术的读者阅读。

更多机器学习相关图书扫码识别直接查看