Five schools of machine learning you don’t know about

Machine learning is an important branch of artificial intelligence that gives computers the ability to learn from data and improve their capabilities without being explicitly programmed. Machine learning has a wide range of applications in various fields, from image recognition and natural language processing to recommendation systems and fraud detection, and it is changing the way we live. There are many different methods and theories in the field of machine learning, among which the five most influential methods are called the "Five Schools of Machine Learning". The five major schools are the symbolic school, the connectionist school, the evolutionary school, the Bayesian school and the analogy school.

1.Symbolism

Symbolism, and Known as symbolism, it emphasizes the use of symbols for logical reasoning and expression of knowledge. This school of thought believes that learning is a process of reverse deduction, using existing knowledge and rules to seek insights from philosophy, psychology and logic. The origins of semiotics can be traced back to ancient times, when early philosophers, logicians, and psychologists studied cognition through the use of symbols. However, truly systematic semiotics began in French culture in the late 19th and early 20th centuries, promoted primarily by a group of writers, artists, and philosophers

• representative Character

Herbert Simon: One of the founders of the semiotic school, he and Allen Newell jointly proposed the universal Concept of Problem Solver (GPS).

Allen Newell: One of the founders of the semiotic school, he and Herbert Simon proposed the universal problem solver ( GPS) concept.

John McCarthy: John McCarthy is one of the pioneers in the field of artificial intelligence and a representative of the semiotic school. He proposed the term "artificial intelligence" in 1956 and developed the LISP programming language, which became an important tool for symbolism research. McCarthy's work mainly focused on logical reasoning and knowledge representation, and he believed that computers could simulate human thinking processes through symbols.

Marvin Lee Minsky: One of the founders of the MIT Artificial Intelligence Laboratory. He proposed the framework theory and made contributions to the field of artificial intelligence. made significant contributions. Marvin Minsky is a leading computer scientist and cognitive scientist. He began studying artificial intelligence in the 1950s and became one of the pioneers in the field. His research focus

• Main algorithm

Inductive Logic Programming (ILP) is a kind of reverse Methods of reasoning. Abverse reasoning usually uses logical reasoning to discover knowledge by extracting general rules from specific examples.

2. Connectionism School

Connectionism, also known as connectionism, is inspired by neural Science and physics, emphasizing reverse engineering of the brain and simulating the structure and function of neural networks. This school of thought believes that intelligence arises through the connections and interactions between large numbers of simple units (neurons). The theory holds that the simulation of connections and interactions between neurons can produce intelligent behavior. This connection and interaction is achieved through connections between simple units (neurons). By adjusting the strength and weight of connections in a neural network, the connections and information transfer between neurons in the human brain can be simulated. One of the main advantages of connectionology is that it allows the generation of intelligence through a large number of simple units Yann LeCun is a brilliant scientist who developed convolutional neural networks and successfully applied them to computer vision tasks such as handwritten digit recognition. LeCun's work has greatly promoted the development of deep learning in practical applications.

• Geoffrey Hinton: A pioneer of deep learning, he proposed important architectures such as convolutional neural networks (CNN) and deep belief networks (DBN).

Yoshua Bengio: A pioneer of deep learning, he proposed important architectures such as the Long Short-Term Memory (LSTM) network.

David Rumelhart: Psychologist, one of the founders of the Parallel Distributed Processing (PDP) model, he proposed the backpropagation algorithm .

Frank Rosenblatt: Psychologist and inventor of the perceptron. He proposed the perceptron learning algorithm.

• Main algorithm

The main algorithm of the connection school is Backpropagation. Backpropagation is an algorithm that updates the weight of a neural network by calculating the gradient of the loss function, which greatly improves the efficiency of training deep neural networks.

3. Evolutionary Computation

Evolutionary Computation is inspired by genetics and evolutionary biology. Learn and optimize by simulating biological evolution processes. The core idea of ​​this school is to use genetic operations such as selection, crossover and mutation to simulate the biological evolution process on the computer to find the optimal solution to the problem.

• Representative figure

John Holland

John Holland is a pioneer in the field of evolutionary computing. He proposed the genetic algorithm (Genetic Algorithm) in the 1960s. Holland's work laid the foundation for evolutionary computation, and his genetic algorithms used natural selection and genetic operations to solve complex optimization problems.

David E.Goldberg

David Goldberg’s research and development on genetic algorithms Made important contributions to applications. His book "Genetic Algorithm" introduces the theory and application of genetic algorithms in detail, which has attracted widespread attention and development in this field.

• Main Algorithm

The main algorithm of the evolutionary school is Genetic Programming (GP). Genetic programming is an algorithm that uses evolutionary computing technology to automatically generate computer programs. It gradually optimizes the program to solve specific problems by simulating the biological evolution process.

4. Bayesian school

Bayesianism is based on statistics and believes that Learning is a process of probabilistic reasoning. This school of thought utilizes Bayes' theorem to perform learning and inference by updating the prior probability distribution.

• Representative figure

Thomas Bayes

Thomas Bayes was a British mathematician. The Bayes theorem he proposed became the basis of Bayesian reasoning. Although Bayes himself was not directly involved in machine learning research, his work was of great significance to the formation and development of the Bayesian school.

Judea Pearl

Judea Pearl works on Bayesian networks and causal inference Made outstanding contributions. His development of Bayesian networks is an important tool that makes probabilistic reasoning in complex systems more efficient and intuitive. Pearl's work has had a profound impact on both artificial intelligence and statistics.

• Main algorithm

The main algorithm of the Bayesian school is Bayesian Inference. Bayesian reasoning makes predictions and decisions by calculating posterior probabilities, and has significant advantages in dealing with uncertainty and complex systems.

5. School of Analogy

The school of analogy (Analogism) learns by extrapolating similarity judgments , influenced by psychology and mathematical optimization. This school emphasizes analogical reasoning from known examples to discover new knowledge and solve problems.

• Representative figure

Vladimir Vapnik

Vladimir Vapnik is one of the important representatives of the analogy school. He and Alexey Chervonenkis jointly proposed the Support Vector Machine (SVM) ). Support vector machine is a supervised learning method based on statistical learning theory and is widely used in classification and regression problems.

Tom Michael Mitchell

Tom Michael Mitchell has made extensive contributions in the field of machine learning , his book "Machine Learning" is an important textbook in this field. Kowalski's research on analogy learning and inductive logic programming provided important theoretical support for the development of the analogy school.

• Main algorithm

#The main algorithm of the analogy school is Support Vector Machine (SVM). Support vector machines implement classification tasks by constructing a hyperplane to maximize the separation between different categories. In high-dimensional data spaces, SVM performs well and is especially suitable for complex pattern recognition problems.

6. Comparison of the five major schools of machine learning

school	Representative	Main idea	##Main algorithm	Application fields
Semiotic School	Herbert Simon, Allen Newell, John McCarthy, Marvin Lee Minsky	Learning is a process of symbol manipulation	Reverse deduction	Knowledge representation, natural language processing
Connection School	Yon LeCun, Jeffrey Hinton, Joshua Bengio, David Rummelhart, Frank Rosenblatt	Learning is a process that simulates the brain’s neural network	Backpropagation	Image recognition, speech recognition, natural language processing
Evolutionary School	John Holland, David Goldberg	Learning is a simulation The process of biological evolution	Genetic algorithm, evolutionary strategy	Robot control and optimization problems Solving
Bayesian	Thomas Bay Yes, Judea Pearl	Learning is a process of probabilistic reasoning	Bayes Theorem	##Spam filtering, medical diagnosis, information retrieval
School of Analogy	Vladimir Vapnik, Tom Michael Mitchell	Learning is a Process by extrapolation of similarity judgments	#Learning algorithm based on analogy	Recommendation system, case reasoning, machine translation

7. Summary

The five major schools of machine learning have their own characteristics, starting from different perspectives and theoretical foundations to solve various complex learning problems. The semiotic school emphasizes logical reasoning and knowledge representation, the connectionist school simulates the structure and function of neural networks, the evolutionary school uses the biological evolution process for optimization, the Bayesian school handles uncertainty through probabilistic reasoning, and the analogical school performs analogical reasoning through similarity judgments. . Each school has its representatives and main algorithms, and their contributions jointly promote the development and progress of the field of machine learning.

Although these five schools have differences in theories and methods, they are not mutually exclusive, but can complement and integrate. In practical applications, researchers often combine multiple methods to deal with complex and changeable problems. With the development of technology and the deepening of interdisciplinary research, machine learning will continue to play an important role in all aspects of artificial intelligence, bringing more innovations and breakthroughs.

The above is the detailed content of Five schools of machine learning you don’t know about. For more information, please follow other related articles on the PHP Chinese website!