1.2.2 Deep Learning

 In 2006, Geoffirey Hinton et al. found that multilayer neural networks can be better trained through layer-by-layer pre-training and achieved a better error rate than SVM on the  MNIST handwritten digital picture data set, turning on the third artificial intelligence revival. In that paper, Geoffrey Hinton first proposed the concept of deep learning. In 2011, Xavier Glorot proposed a Rectified Linear Unit (ReLU) activation function, which is one of the most widely used activation functions now. In 2012, Alex Krizhevsky propeosed an eight-layer deep neural network AlexNet, which used the ReLU activation functio nand Dropout technology to prevent overfitting. At the same time, it abandoned the layer-byt-layer pre-training method and directly trained the network on two NVIDIA GTX580 GPUs. AlexNet won the first place in the ILSVRC-2012 picture recognition competition, showing a stunning 10.9% reduction in the top-5 error rate compared with the second place.

Since the AlexHNet model was developed, various models have been published successively, including VGG series models increase the number of layers in the network to hundreds or even thousands while manintaining the smae or even better performance, which is the most representative model of deep learning.

In addition to the amazing results in supervised learning, huge achievements have also been made in unsupervised learning and reinforcement learning. In 2014, Ian Goodfellow proposed generative adversarial networks(GANs), which learned the true distribution of samples through adversarial training to generate samples with higher approximation. Since then, a large number of GAN models have been proposed. The latest image generation models can generate images that reach a degree of fidelity hard to discern from the naked eye. In 2016, DeepMind applied deep neural networks to the field of reinforcement learning and proposed the DQN algorithm, which achieved a level comparable to or even higher than that of humans in 49 games in the Atarigame platform. In the field of Go, AlphaGo and AlphaGo Zero intelligent programs from Deep Mind have successively defeated hyuman top Go players Li Shishi, Ke jie, etc. In the multi-agent collaboration Dota 2 game platform, OpenAI five intelligent programs developed by OpenAI defeated the T18 champion team OG in a restricted game environment, showing a large number of professional high-level intelligent operations. Figure 1-9 lists the major time points between 2006 and 2019 for AI development.

1.2.1 Shallow Neural Netorks

 In 1943, psychologist Warrent McCulloch and logician Walter Pitts proposed the earliest mathematical model of neuraons based on tghe structure of biological neuraons, called MP neuraon models after their last name initials. The model f(x)=h(g(x)), where g(x)=iXi, Xi∈{0,1}, takes values from g(x) to predict output values as shown in Figure 1-4. If g(x) >=0, output is 1; if g(x) < 0, output is 0. The MO neuraon models have no learning ability and can onlyu complete fixed logic judgments.

In 1958, American psychologist Frank Rosenblatt proposed the first neuron model that can automatically learn weights, called perceptron. As xshown in Figure 1-5, the error between the output value 0 and the true value y is used to adjust the weights of the neuraons {w1, w2, w3...wn}. Frank Rosenblatt then implemented the perceptron model based on the "Mark 1 perceptron" hardware. As shown in Figures 1-7 and 1-7, the input is an image sensor with 400 pixels, and the output has eight nodes. It can successfully identify some English letters. It is generally believed that 1943-1969 is the first prosperous period of artificial intelligence development.

In 1969, the American scientist Marvin Minsky and others pointed out the main flaw of linear models such as perceptrons in the book Perceptrons. They found that perceptrons cannot handle simple linear inseparable problems such as XOR. This directly led to the trough period of perceptron-related research on neural networks. It is generally considered that 1969-1982 was the first winter of artificial intelligence.

Although it was in the trough period of AI, there were still many significant studies published one after another. The most important one is the backpropagation(BP) algorithm, which is still the core foundation of modern deep learning algorithms. In fact, the mathematical idea of the BP algorithm has been derived as early as the 1960s, but it had not been applied to neural networks at that time. In 1974, American scientist Paul Werbos first proposed that the  BP algorithm can be applied to neural networks in his doctoral dissertation. Unfortunately, this result has not received enough attention. In 1986, David Rumelhard et al. published a paper using the BP algorithm for feature learning in Nature, Since then, the BP algorithm started gaining widespread attention.

In 1982, with the introduction of John Hopfield[s cyclically connected Hopfield network, the second wave of artificial intelligence renaissance was started from 1982 to 1995. During this period, convolutional neural networks, recurrent neural networks, and backpropatation algorithms were developed one after another. In 1986, David Rumelhart, Geoffreey Hiton, and other applied the BP algorithm to multilayer perceptrons, in 1989, Yann LeCun and other applied the BP algorithm to handwritten digital image recognition and acghieved great success, which is known as LeNet. The LeNet system was successfully commericalled in zip code recognition, bank check recognition, and many other systems. In 1997, one of the most widely used recurrent neural network variants, Long ShortTerm Memory(LTSM), was proposed by Jurgen Schmidhuber. In the same year, a bidrectional recurrent neural network was also proposed.

Unfortunately, the study of neural networks has graduallyu entered a though with the rise of traditional machine learning algorithms represented by support vector machines(SVNs), which is known as the second winder of artificial intelligence. Suppport vaector amchines have a rigoroutstheoretical founda5tion, requre a small number of training samples,a nd also have good generalization capabilities. In contrast, neural networks lack theorerical foundation and are hard to interpret. Deep networks are difficult to train, and the performance is normal. Figure 1-8 shows the significant time of AI developemnt between 1943and 2006

1.2 The History of Neural Networks

 We divide the development of neural networks into shallow neural networks stages and deep learning stages, with 2006 as the dividing point. Before 2006, deep learning developed under the name of neural networks and experienced two ups and two downs. In 2006, Geoffrey Hinton first named deep neural networks as deep learning, which started its third revival.

1.1.3 Neural Networks and Deep Learning

 Neural network algorithms are a class of algorithms that learn from data based on neyural networks. They still belong to the category of machine learning. Due to the limitation of computing power and data volume, early neural networks were shallow, usually with around one to four layers. Therefore, the network expression ability was limited. With the improvement of computing power and the arrival of the big data era, highly parallelized graophics processing units(GPUs) and massive data make training of large-scale neural networks possible.

In 2006, Geoffrey Hinton first proposed the concept of deep learning. In 2012, AlexNet, and eight-layer deep nerural network, was released and achieved huge performance imprevements in the image recongnition competition. Since then, neural network models with dozens, hundreds, and even thousands of layers have been developed successively, showing strong learning ability. Algorithms implemented using deep neural networks are generallty referred to as deep learning models. In essence, neural networks and deep learning can be considered the same.

Let's simply compare deep learning with other algorithms. As shown in Figure 103, rule-baed systems usually write explicit logic, which is generally designed for specific tasks and is not suitable for other tasks. Traditional machine learning algorithms artificially design feature detection methods with certain generality, such as SIFT nad HOG features. These featureas are suitable for a certain type of tasks and have certain generality. but the performance highly depends onb how to desing those features. The emergence of neural hetworks ahs made it possible for omputers to design those features automatically through nerual nwtworks without human intervention. Shallow nerual networks typically have limited feature extraction capability, while deep neural networks are capable of extracting high-level, abstract features and have better performance.

1.1.2 Machine Learning

 Machie learning can be divided into supervised learning, unsupervised learning, and reinforcement learning, as shown in Figure 1-2.

Supervised Learning. The supervised learning data set contains samples x and sample labels y. The algorithm needs to learn the mapping relationship f0: x->y, where f0 represents the model function and  are the parameters of the model. During training, the model parameters  are optimized by  minimizing errors between the model prediction and the real value y, so that the model can have more accurate prediction. Common supervised learning models include linear regression, logistic regression, support vector machines (SVMs), and random forests.

Unsupervised Learning. Collecting labeled data is often more expensive. For a sample-only data set, the algorithm needs to discover the modalities of the data itself. This kind of algorithm is called unsupervised learning. One type of algorithm in unsupervised learning uses itself as a supervised signal, that is, f:x->x, which is known as self-supervised learning. During training, parameters are optimized by minimizing the error between the model's predicted value f(x) and itself x, common unsupervised learning algorithms include self-encoders and generative adversarial networks(GANs)

Reinforcement Learning. This is a type of algorithm that learns strategies for solving problems by  interacting with the environment, Unlike supervised and unsupervised learning, reinforcement learning problems do not have a clear "correct" action supervision signal. The algorithm needs to interact with the environment to obtain a lagging reward signal from the environmental feedback. Therefore, it is not possible to calculate the errors between model Reinforcement Learning prediction and "correct values" to optimize the network directly. Common reinforcement learning algorithms are Deep Q-Networks(DQNs) and Proximal Policy Optimizaiton(PPO).

1.1.1 Artificial Intelligence Explained

 AI is a technology that allows machines to acquire intelligenct and inferential mechanisms like humans. this concept first appeared at the Dartmouth Conference in 1956. This is a very challenging task. At present, human beings cannot yet have a comprehensive and scientific understanding of the working mechanism of the human brain. It is undoubtedly more difficult to make intelligent machines that can reach the level of the human brain. With that being said, machines that archive similar to or even supass huyman intelligence in some way have been proven to be feasible.

How to relize AI is very broad question. The development of AI has mainly gone thorough three stages, and each stage represent the exploration footprint of the human trying to realize AU from differenct angles. In the early stage, people tried to develop intelligent systems by suymmarizing and generalizing some logical rules and implementing them in the form of computer programs. But such explicit rules are often too simple and are difficult to be used to express complex and abstract concepts and rules. This stage is called the inference period.

In the 1970s, scientists tried to implement AI though knowledge database and reasoning. They built a large and complex expert system to simulate the intelligence level of human experts. One of the biggest difficulties with theses explicitly specified rules is that many complex, abstract concepts cannot be implemented in concrete code. For example, the process of human recognition of pictures and understanding of languages cannot be simulated by established rules at all. To solve such problems, a research discipline that allowed machines to automatically learn rules from data, known as machine learning, was born, Machine learning become a popular subject in AI in the 1980s, This is the second stage.

In machine learning, there is a directiion to learn complex, abstract logic through neural networks, Research on the direction of neural networks has experienced two ups and downs. Since 2012, the applications of deep neural network technology have made major breakthroughs in fields like computer vision, natural language processing(NLP), and robotics. Some tasks have even surpassed the level of human intelligence. This is the third revival of AI. Deep neural networks eventually have a new name -  deep learning. Generally speaking, the sessential difference between neural networks and deep learning is not large. Deep learning refers to models or algorithms based on deep neural networks. The relationship between artificial intelligence, machine learning, neural networks, and deep learning is shown in Figure 1-1.

1.1 Artificial Intelligence in Action

 Information technology is the third industrial revolution in human history. The popularity of computers, the Internet, and smart home technology has greatly facilitated people's daily lives. Through programming, humans can hand over the interaction logic designed in advance to the machine to execute repeatedly and quickly, thereby freeing humans from simple and tedious repetitive labor. However, for tasks that require a high level of intelligence, such as face recognition, chart robots, and autonomous driving, it is difficult to design clear logic rules. Therefore, traditional programming methods are powerless to those kinds of tasks, whereas artificial intelligence(AI), as the key technology to solve this kind of problem, is very promising.

Wigh the rise of deep learning algorithms, AI has achieved or even surpassed hymanlike intelligence on some tasks. For example, the AlphaGo program has defeated Ke Jie, one of the stongest human Go players, and OpenAI Five has beaten the champion team OG on the Dota2 game. In the meantime, practical technologies such as face recognition, intelligent speech, and machine translation have entered people's daily lives. Now our lives are actully surrounded by AI. Although the current level of intelligence that can be reached is still a long way from artificial general intelligence(AGI), we still firmly believe that the era of AI has arrived.

Next, we will introduce the concepts of AI, machine learning, and deep learning, as well as the connecctions and differences between them.