One, the introduction
What is artificial intelligence?
If you’re a fan of new technology or an engineering student, you’ve probably heard of artificial intelligence (AI). Here’s baidu’s definition of what it is:
Artificial Intelligence, or AI for short. It is a new technical science to research and develop the theory, method, technology and application system for simulating, extending and expanding human intelligence. Artificial intelligence is a branch of computer science that seeks to understand the nature of intelligence and produce a new type of intelligent machine that can react in a manner similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. Since the birth of artificial intelligence, the theory and technology are increasingly mature, and the application field is also expanding. It can be imagined that the future scientific and technological products brought by artificial intelligence will be the “container” of human wisdom. Artificial intelligence can simulate the information process of human consciousness and thinking. Artificial intelligence is not human intelligence, but it can think like a human and possibly exceed human intelligence. Artificial intelligence is a challenging science, and those who do it must have knowledge of computers, psychology and philosophy. Artificial intelligence is a very broad science, which consists of different fields, such as machine learning, computer vision and so on. In general, one of the main goals of artificial intelligence research is to enable machines to perform some complex tasks that normally require human intelligence. However, different generations and different people have different understandings of this “complex work”. In December 2017, AI was selected as one of the top 10 Buzzwords in Chinese media in 2017. —- from Baidu
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 -> human confrontation and cooperation ability (6) machine learning -> Human ability to learn
2. Practical application
Machine vision, Fingerprint recognition, Face recognition, retina recognition, iris recognition, palmprint recognition, expert systems, automatic planning, intelligent search, theorem proving, game, automatic programming, intelligent control, robotics, language and image understanding, genetic programming, etc.
3. Subject category
Artificial intelligence is a borderline subject, belonging to the intersection of natural science and social science.
4. Involve disciplines
Philosophy and cognitive science, Mathematics, neurophysiology, psychology, computer Science, information theory, cybernetics, uncertainty
5. Research scope
Natural language processing, knowledge representation, intelligent search, reasoning, programming, machine learning, knowledge acquisition, combinatorial scheduling problems, perceptual problems, pattern recognition, logical programming, soft computing, management of inaccuracies and uncertainties, Artificial life, neural networks, complex systems, Genetic algorithms
Consciousness and artificial intelligence
In essence, artificial intelligence is the simulation of the information process of human thinking.
There are two ways to simulate people’s thinking. One is to simulate the structure of human brain and make a machine like human brain. Second, functional simulation, temporarily put aside the internal structure of the human brain, and from its functional process simulation. The emergence of modern electronic computer is the simulation of the thinking function of human brain and the information process of human brain.
Weak ai now continuously rapid development, especially after the economic crisis in 2008, Japan Europe wants to use the robot to realize industrialization, such as industrial robots to the faster than any time before the development, more led to weak and related industries in the field of artificial intelligence breakthrough unceasingly, many must choose and employ persons to do work now can use the robot to realize.
However, strong artificial intelligence is at a bottleneck for the time being, which requires the efforts of scientists and human beings.
Second, the introduction
Back to the main topic of our article, the way to achieve artificial intelligence (AI) : machine learning, deep learning.
1. Machine learning: an approach to artificial intelligence
At its most basic, machine learning uses algorithms to parse data, learn from it, and then make decisions and predictions about real-world events. Unlike traditional software programs that are hard-coded to solve specific tasks, machine learning takes large amounts of data and “trains” them, using algorithms to learn from the data how to perform a task. To take a simple example, when we browse online shopping malls, we often see the information of product recommendation. It’s the store’s way of identifying what you’re really interested in and willing to buy, based on your past shopping history and lengthy collection lists. Such a decision-making model can help the mall offer suggestions to customers and encourage product consumption. Machine learning is directly derived from the early field of artificial intelligence. Traditional algorithms include decision tree, clustering, Bayesian classification, support vector machine, EM, Adaboost and so on. In terms of learning methods, machine learning algorithms can be divided into supervised learning (such as classification problems), unsupervised learning (such as clustering problems), semi-supervised learning, integrated learning, deep learning and reinforcement learning. The application of traditional machine learning algorithms in fingerprint recognition, face detection based on Haar, object detection based on HoG feature and other fields has basically reached the requirements of commercialization or the commercialization level of specific scenes, but every step is extremely difficult until the emergence of deep learning algorithm.
2. Deep learning: a technique for implementing machine learning
Deep learning is not an independent learning method, and it also uses supervised and unsupervised learning methods to train deep neural networks. However, due to the rapid development of this field in recent years, some unique learning methods have been proposed (such as residual network), so more and more people regard it as a learning method alone. The original deep learning is a learning process that uses deep neural networks to solve feature expression. Deep neural network itself is not a new concept, which can be roughly understood as a neural network structure containing multiple hidden layers. In order to improve the training effect of deep neural network, the neuron connection method and activation function are adjusted accordingly. In fact, a lot of ideas had been put forward in the early years, but due to the lack of training data and backward computing ability, the final effect was not satisfactory. Deep learning performs tasks so devastatingly that it makes all sorts of machine-assisted functions seem possible. Driverless cars, preventive health care, even better movie recommendations are all in sight, or within reach.
3. The differences and connections among the three
Machine learning is a method to achieve artificial intelligence, and deep learning is a technology to achieve machine learning. Let’s visualize the relationship between them using the simplest method, concentric circles.
There is now a widespread sense in the industry that deep learning may eventually make all other machine learning algorithms obsolete. This awareness is mainly due to the fact that the current application of deep learning in computer vision and natural language processing far exceeds traditional machine learning methods, and the media has greatly exaggerated the coverage of deep learning. Deep learning, as the hottest machine learning method at present, is not meant to be the end of machine learning. There are at least the following problems:
- The deep learning model needs a large amount of training data to show the magic effect, but in real life, it often encounters small sample problems. In this case, the deep learning method cannot be started, and the traditional machine learning method can deal with them.
- Some domains can be well addressed by traditional simple machine learning methods, without the need for complex deep learning methods.
- The thought of deep learning, comes from the inspiration of the human brain, but by no means the simulation of the human brain, for example, for a three or four years old children, after watching a bike to see even look completely different bicycle, children are nine times out of ten can make that judgment is a bicycle, that is to say, human learning process often don’t need a large training data, And the current deep learning method is obviously not a simulation of the human brain.
Introduction to three,
As a fan of new technology, I believe that everyone like me wants to learn new technology, but has no idea how to get started. Here are some recommendations for you:
1, 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.
Deep Learning: “Flower book”, also regarded as the classic book of deep learning.
Machine Learning in action.
Hands-on deep learning.
“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.
2, video
- Stanford CS229 Machine Learning by Andrew Ng
Video 2009: Stanford class video, full of content, but on the blackboard, and some unnecessary classroom interaction, easy to distract.
Video 2014: This is Ng’s coursera lecture, one topic at a time, more concise and clear.
- 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.
- Machine Learning and Neural Networks -Hinton:
I’m afraid there’s no better way to talk about neural networks than grandpa. Hinton’s mind is very deep, in other words not very easy to understand, but that doesn’t mean it’s a very good course.
- Stanford CS231- Deep Learning Computer Vision – Feifei Li:
Computer vision is definitely riding the wave of deep learning, and it’s a classic introductory video for computer vision, which will introduce all kinds of convolutional neural networks.
- 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.
- 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!
- Machine Learning – Whiteboard Derivation series:
The teaching videos launched by station B are well organized and focused, and each chapter only lasts about 20 minutes. The main language speed of up is slow and there is no problem to watch it at 1.5 or 2 times speed, which is very suitable for beginners.
Four, advice,
No matter how much you learn, it is better to do it yourself, whether it is a project or a competition.