1. Basic Concepts
1. What is artificial intelligence
The Concept of Artificial Intelligence: Machines Simulate Human Consciousness and Thinking
Important people : Alan Mathison Turing _ _
Character brief introduction: June 23 , 1912 - June 7 , 1954 , British mathematician and logician , known as
The father of computer science, the father of artificial intelligence.
Related events: (1) In 1950, in the paper "Can machines think? , proposed the Turing test, a method for
Test method to determine whether a machine has intelligence: the questioner and the answerer are separated, and the questioner passes through some device
(like a keyboard) to ask random questions to the machine. Test multiple times, if more than 30% of the questioners think the question is answered
is a human rather than a machine, then the machine passes the test and has artificial intelligence. artificial intelligence
The concept of: "Modeling human consciousness and thinking with machines".
(2) Turing predicted in the paper: In 2000, there will be machines with artificial intelligence that pass the Turing test.
However, it wasn't until June 2014 that the chat app at the University of Reading in the UK managed to impersonate a 13-year-old boy through
Turing test. This event came 14 years later than Turing predicted.
(3) In the cover news report of Science magazine in November 2015, robots have been able to
A character in the text that has been passed, writing a character of the same style, indicating that the machine has the ability to quickly learn
The creative ability of raw words.
2. What is machine learning
The concept of machine learning: Machine learning is a statistical method in which a computer uses existing data to obtain a certain model.
model, and then use this model to predict the results .
Features: With the increase of experience, the effect will become better.
Simple model example: decision tree model
The difference between machine learning and traditional computer computing: traditional computers are based on the von Neumann structure, and the instructions are pre-
storage. When running, the CPU reads the instructions line by line from the memory and executes the prearranged line by line step by step.
instruction. Its characteristic is that the output result is determined, because what to do first and what to do after has been written in the instruction in advance
in.
Three elements of machine learning: data, algorithms, and computing power
3. What is deep learning
The concept of deep learning: deep neural network, derived from the study of the structure of biological brain neurons .
Human brain neural network: As people grow, the brain neural network gradually becomes thicker and stronger.
Neurons in biology: There are many tributaries on the left side of the figure below, which are called in biology
dendrites. Dendrites have the function of receiving stimuli and transmitting impulses to the cell body, which is the input of neurons. these trees
The synapses converge in the nucleus and exit along an axon. The main function of the axon is to transmit nerve impulses from the cell body to the
Other neurons, are the outputs of neurons. The human brain is made up of 86 billion such neurons, all
Consciousness of thinking is realized by taking it as the basic unit and connecting it into a network.
Neuron Models in Computers: In 1943 , psychologist McCulloch and mathematician Pitts referenced
The structure of biological neurons, published an abstract neuron model MP . A neuron model is an input that contains,
Models that output and compute functions. The input can be analogized to the dendrite of a neuron, and the output can be analogized to a neuron
The axon, the calculation can be analogized to the nucleus.
4. Artificial Intelligence Vs Machine Learning Vs Deep Learning
Artificial intelligence is the use of machines to simulate human consciousness and thinking.
Machine learning, on the other hand, is a method of realizing artificial intelligence and is a subset of artificial intelligence.
Deep learning is a deep neural network, an implementation method of machine learning, and a subset of machine learning.
Second, the development history of neural networks (three ups and two downs)
The first rise: In 1958 , people connected two layers of neurons end-to-end to form a single-layer neural network, which is called sensory neural network.
Know the machine . The perceptron became the first artificial neural network that could learn. Triggered the first of neural network research
rise again.
The first cold winter: In 1969 , Minsky , an authoritative scholar in this field, used mathematical formulas to prove that only single
The perceptron of the layer neural network cannot classify the XOR logic. Minsky also pointed out that in order to solve the XOR, it is possible to
To solve the problem, it is necessary to extend the single-layer neural network to two or more layers. However, in that era, the
Computational power cannot support this amount of computation. A perceptron with only one layer of computing units, exposing his genius
However, due to the defects, neural network research has entered the first cold winter.
The second rise: In 1986 , Hinton et al. proposed the backpropagation method, which effectively solved the two-layer neural network.
The computing power of the network. Triggered the second rise of neural network research.
The second winter: In 1995 , the support vector machine was born. Support vector machines can eliminate the need for neural networks to adjust
The lack of parameters also avoids the problem of local optima in neural networks. Beat the neural network in one fell swoop
At that time, the mainstream algorithm in the field of artificial intelligence made the neural network enter his second winter.
The third rise: In 2006 , the emergence of deep neural networks, in 2012 , the convolutional neural network in image recognition
The amazing performance in other fields has triggered another rise in neural network research.
3. Typical applications of machine learning
1. Application areas
Computer Vision, Speech Recognition, Natural Language Processing
2. Mainstream applications :
( 1 ) Prediction (predicting continuous data)
( 2 ) Classification (classify discrete data)
4. Course Summary
1. Machine learning means that on the task T , with the increase of experience E , the effect P increases.
2. The process of machine learning is to generate a model through the input of a large amount of data, and then use the generated
Models that make predictions about outcomes.
3. The huge neural network is based on the neuron structure. The input is multiplied by the weight, then summed, and then non-linear.
process of sexual function.