\

Ingenious OR ingenious

By Howard

Editor’s note:

Getting into machine learning and deep learning is not an easy task. There is a lot of knowledge to learn, and beginners are often at a loss. Now we bring you a pure dry article to help you get started on machine learning and deep learning.

What is artificial intelligence?

According to the definition of UCLA professor Zhu Songchun, AI can be roughly divided into the following six categories:

(1) Computer vision -> human visual ability

(2) Natural language processing -> human language ability

(3) Speech recognition and generation -> human listening and speaking ability

(4) Robotics -> Human motor ability and motor intelligence

(5) Game and cooperation -> The ability of human confrontation and cooperation

(6) Machine learning -> human learning ability

You can see that the previous technique was to replace the old tool with the new one. Artificial intelligence is different, its goal is to imitate human intelligence, replace human intelligence, surpass human intelligence.

So what exactly is AI at this stage?

I think the core concept of AI is still statistical inference or function fitting. Speaking is to find a function or a mapping so that a sound wave maps to language, a picture of a cat maps to a cat. When you say hi to Siri, it automatically maps to hello ~

Learn deep Learning in one day – Li Hongyi

Introduction to artificial intelligence short book list

I recommend some books to you and remind you of some pits. Books can be roughly divided into three categories: popular science books, machine learning algorithm books, and programming books.

1. Popular Science books:

The Beauty of Mathematics: The mathematics of natural language processing and search engines

Top of the Wave: Tells the development of IT industry and the rise and fall of IT companies in Silicon Valley

Hackers and Painters: This book is a collection of essays by Graham, the founder of Silicon Valley

Interest is the best teacher. Reading popular science books helps to accumulate interest and have a general understanding of the field of artificial intelligence.

2. Machine Learning Algorithm Books:

Statistical Learning Methods: a classic textbook of Dr. Li Hang. Describe machine learning algorithms in the most concise language, a must-read book for career AI

Machine Learning: Professor Zhou Zhihua’s watermelon book. Statistical learning methods cover too narrow, with watermelon book to expand the width.

Python Machine Learning and Implementation: Easy to get started with. The learning curve is smooth. If you are tired of reading the theory books, you can type code with this book and get a general understanding of Kaggle.

“Collective programming wisdom” : there are various algorithms to achieve the code, with the theory of the book, can be more in-depth understanding of the algorithm.

PRML: Machine learning classics, Bayesian classics.

Neural Networks and Deep Learning: Qiu’s Open Source book (nndl.github. IO /)

3. Programming Books:

Liao Xuefeng Python Tutorial: The best Introduction to Python, www.liaoxuefeng.com

Smooth Python: The best advanced tutorial for Python

Python for Data Analysis is written by Python Pandas. It is used to analyze Data in Pandas

4. Some pits:

Tensorflow In Action: It’s really better to read the official tutorial or CS 20SI at Stanford

Deep Learning: Goodfellow’s masterpiece, but really not suitable for beginners, suitable for advanced

5. Interview Books:

“100 sides machine learning” : commonly known as the gourd book, has been the way of asking questions summarized in machine learning interview points, surface algorithm suggested to prepare a book

5. “Finger offer” : Read books because many interviewers get questions from them

Introduction to artificial intelligence video

Introductory Stage:

  • Stanford CS229 Machine Learning by Andrew Ng

    The first is CS229 by Ng. Ng is really a natural good teacher with clear lectures, clear rules, difficult issues and a smooth learning curve

Video 2009: Stanford class video, full of content, but on the blackboard, and some unnecessary classroom interaction, easy to distract

I know you’re all hands on deck, and the link is ready.

Stanford Open Course: Machine Learning Course

(link.zhihu.com/?target=htt…

Video 2014: This is Ng’s coursera lecture, one topic at a time, more concise and clear

Link: Machine Learning, Machine Learning) – Wu En da (Andrew Ng) | Stanford university course CS229 (2014).

(www.bilibili.com/video/av990…

  • Fundamentals of Machine Learning and Machine Learning Techniques – Lin Xuantian:

One problem with Ng’s courses is that they are taught in English, so some students are afraid of English and shrink back.

In this case, the machine learning foundation and techniques of Lin Xuantian from Taiwan University are a very good choice.

This course is a little bit deeper, a little bit more mathematical, and it’s going to cover some very basic machine learning theories like VC dimension, KKT conditions, etc

My advice is to go through what you don’t understand at first and then regurgitate it later

Links: fundamentals of machine learning, full edition _ lectures • open courses _ technology _bilibili_ bilibili

(www.bilibili.com/video/av429…

Advanced stage:

  • Machine Learning and Neural Networks -Hinton:

There’s no better way to talk about neural networks than grandpa, and Hinton’s mind is very deep, in other words, not very easy to understand

But that doesn’t take away from the fact that it’s an excellent course

Link: Hinton Machine Learning and Neural Network Chinese course – netease Cloud Classroom

(study.163.com/course/intr…

  • Stanford CS231- Deep Learning Computer Vision – Feifei Li:

Computer vision is undoubtedly the wave of deep learning, do computer vision direction classic introductory video

There are all kinds of convolutional neural networks

Link: Stanford Fei-fei Li – Deep learning computer vision – netease Cloud Classroom

(study.163.com/course/frie…

  • Stanford CS224- Deep Learning natural Language Processing -Chris Manning

Another important area of ARTIFICIAL intelligence is natural language processing, and the classic primer on this is CS224

Stanford Season 2017 CS224n Deep Learning Natural Language Processing Course by Chris Manning & Richard Socher

(www.bilibili.com/video/av133…

CS224n Advanced Natural Language Processing course at Stanford

(www.bilibili.com/video/av413…

  • Machine Learning — Li Hongyi

The above video is classic, but it is taught in English, which makes a lot of babies upset, but it’s ok, come to Taiwan University’s deep learning!

Li Hongyi Machine Learning (2017, Autumn, Taiwan University) Mandarin

Machine Learning – Li Hongyi (2019) Machine Learning

(www.bilibili.com/video/av359…

  • Machine learning – Whiteboard derivation series

Each chapter only lasts about 20 minutes, and the up main language speed is slow. 1.5 or 2 times speed is not a problem. It is very suitable for beginners

Links: bi li bi li (゜ ゜ つ ロ cheers ~ Bilibili

(space.bilibili.com/97068901/ch…

See here, you will find that, in fact, as long as you want to learn, there are a lot of videos, the Internet makes the acquisition of knowledge is no longer difficult, really hinder you only your determination and perseverance.

Last but not least, don’t watch zhang Zhihua’s video on statistical machine learning. His level is very high, but the formulas on the screen in the video remind me of my fear of being dominated by various theorems in college. Chinese university classes have been in the sense of persuasion, remind you can not learn, you do not understand, your IQ is not enough. However, Ng’s class does not. He thinks there are no students who can’t learn, only teachers who can’t teach

Introduction to artificial intelligence learning methods

One of the pitfalls of machine learning beginners is to get caught up in the pursuit of big algorithms. Can I use deep learning to solve this problem? Boosting is a boosting algorithm that I need to do some model fusion with. I’ve always had the view that “talking about algorithms outside of business and data is pointless”.

In fact, according to our learning experience, starting from a data source, even if we use the most traditional machine learning algorithm that has been applied for many years, we should first go through the whole process of machine learning completely, constantly try various algorithms to dig into the value of these data, and thoroughly understand the data, features and algorithms in the application process. Truly accumulating project experience is the fastest and most reliable learning path.

So how do you get data and projects? A shortcut is to actively participate in various data mining competitions at home and abroad, download the data directly, and continue to optimize according to the requirements of the competition to accumulate experience. Kaggle, DataCastle and Ali Tianchi are great platforms where you can get real data and learn and compete with data scientists. It’s fun to try to complete the competition using all the knowledge you’ve learned. Discussions with other data scientists can broaden your horizons and give you a deeper understanding of machine learning algorithms.

Interestingly, some platforms, such as Ali Tianchi Competition, even provide all components from data processing, model training, model evaluation, visualization and model fusion enhancement. All you need to do is participate in the competition, obtain data, and then use these components to implement your idea. Cut the crap and get to the point. If you think machine learning is hard, you’re not opening it the right way.

Machine learning may seem difficult, but there is a universal learning path for beginners. As I’ve seen in previous columns, there are plenty of excellent get-to-know-you materials that can lower the bar for learning and stimulate the fun of learning.

To put it simply, the learning methods are as follows:

  1. Programming skills
  2. Machine learning knowledge —-> hands-on training code —-> data science competition —-> practical project experience
  3. Mathematical basis

Machine learning is a field that combines theoretical algorithms with computer engineering techniques. You need a solid theoretical foundation to help you analyze data, as well as engineering skills to develop models and deploy services. Therefore, it requires [programming skills] [machine learning knowledge] [mathematical fundamentals] three armies to advance together in order to finally win the fruits of victory.

There are also three types of people who switch to AI. One is a programmer with good engineering experience. One is statistics, mathematics, electronic communication, with a solid theoretical foundation; There is also a kind of neither rich programming experience nor solid theoretical basis, such as we learn materials… These three types of students need to strengthen the part of AI is different.

1. Programming skills – Python

Life is short, I use python!

With Google, Facebook and the rest of the world giving the go-ahead, python is one of the hottest names in artificial intelligence. Python’s tool library is fairly extensive, ranging from data retrieval to data cleansing and machine learning algorithms. More comprehensive than R.

In addition to mastering python’s own syntax, you should also focus on the following libraries:

Pandas: A powerful tool for manipulating data in tables and data cleaning and preprocessing.

Numpy: numerical computation library, fast don’t don’t.

Matplotlib: Data visualization tool that mimics MATLAB.

Scikit-learn: Encapsulates a super good machine learning library, and some simple algorithms are not too easy to use.

Ipython Notebook: A notebook for data scientists and algorithm engineers, highly recommended.

2. Math basics

Calculus: Calculus is one of the core knowledge in machine learning, whether in gradient descent method to calculate the gradient or the derivation of error transmission in back propagation need to use calculus.

Linear algebra: a large number of calculations in neural networks are matrix multiplication, which requires knowledge of linear algebra. The inner product operation is also used to calculate cosine similarity of vectors. In addition, various matrix decomposition methods also appear in principal component analysis and singular value decomposition.

Probability theory and Statistics: Broadly speaking, the core of machine learning is statistical inference. Therefore, many machine learning giants are masters of statistics, such as Michael Jordan, Yang Lekun, Sinton and so on. Bayesian formula, hidden Markov model and so on are widely used in machine learning.

Admittedly, math is very important, but I suggest, with undergraduate mathematical foundation, you don’t spend too much time to brush a math book, this is the opposite, the best way is taking the huanglong learning machine learning algorithms, to don’t understand the place to add the corresponding mathematical knowledge, Stanford tutorial, there are a lot of math supplementary material, Most of the time it is enough to read the supplementary material.

3. Project experience

One misconception: Many newbies start with complex deep learning models and advanced algorithms such as AlexNet and ResNet. Call all the apis of TensorFlow and Keras without knowing what you’re doing.

One idea: Algorithms divorced from real business and data are castles in the air.

One path: According to my learning experience, the fastest and most reliable learning method is to use the most traditional algorithm to go through the whole process from data cleaning to feature engineering, and to constantly compare and try various algorithms to get the features and algorithms thoroughly.

Two projects: Foreign Kaggle and Ali Cloud Tianchi are good ways to gain project experience. My recommendation is that everyone starting out in machine learning should take two programs. A traditional machine learning application scenario project, such as Ali mobile recommendation algorithm. Familiar with logistic regression, support vector machine and gradient enhanced decision tree algorithms through traditional application scenarios. A deep learning application scenario project, such as lung cancer identification and diagnosis, through deep learning application scenarios to get familiar with the advantages and application scenarios of various algorithms of deep learning.

That’s all. Let’s go for it! Beautiful angel beckons you from afar, brave boy, go and work miracles!

Note: the menu of the official account includes an AI cheat sheet, which is very suitable for learning on the commute.

Machine learning Online Manual Deep Learning online Manual AI Basic Download (Part 1)4500+ user ID:92416895), please reply to knowledge PlanetCopy the code

Like articles, click Looking at the