Machine Learning
Definition: The process by which computers acquire knowledge through automatic analysis and synthesis of data, facts, or their own experiences.
Discipline: Computer Science Technology _ Artificial Intelligence _ Machine Learning
Related Terms: Artificial Intelligence, Big Data, Pattern Recognition, Deep Learning
Image Source: Visual China
[Further Reading]
Machine learning is a branch of artificial intelligence that studies human learning behavior. It draws on perspectives from cognitive science, biology, philosophy, statistics, information theory, control theory, and computational complexity, exploring the laws of human cognition and learning processes through basic methods such as induction, generalization, specialization, and analogy, establishing various algorithms that can automatically improve through experience, enabling computer systems to learn specific knowledge and skills autonomously, and creating task-oriented learning systems with specific applications.
The core of machine learning is that machines use algorithms to analyze vast amounts of data, learning from the data to mine the potential relationships within it and train an effective model for decision-making or prediction. Currently, machine learning is divided into two basic types: supervised learning and unsupervised learning. Supervised learning uses labeled data as the final learning objective, typically yielding good results, but acquiring labeled data is costly; unsupervised learning is akin to self-learning or self-service learning, making it easier to utilize more data, potentially discovering more prior knowledge of patterns existing in the data (sometimes surpassing manually labeled pattern information), but with lower learning efficiency. Both share the commonality of establishing mathematical models to solve optimization problems, usually without a perfect solution. Generally, when faced with a large amount of data but lacking understanding, we can use unsupervised learning.
Machine learning can be traced back to research on artificial neural networks. In 1943, neurologist Warren McCulloch and mathematician Walter Pitts proposed a hierarchical model of neural networks, establishing the theoretical foundation for the computational model of neural networks, thereby laying the groundwork for the development of machine learning. In 1950, the “father of artificial intelligence” Alan Turing proposed the famous “Turing Test,” making artificial intelligence an important research topic in the field of computer science. In 1957, Cornell University professor Frank Rosenblatt introduced the concept of the Perceptron and precisely defined the mathematical model of self-organizing and self-learning neural networks for the first time with an algorithm, designing the first computer neural network, which became the progenitor of neural network models in machine learning.
For over 30 years, a topic in the field of machine learning called “deep learning” has received widespread attention, achieving significant progress in areas such as speech, image, natural language, and online advertising. The famous AlphaGo (Go robot) once engaged in a Go battle with top Korean player Lee Sedol, ultimately winning 4 to 1. This fact sufficiently demonstrates the strong capabilities of machine learning, prompting people to envision a bright future for deep machine learning. In the near future, machine learning is bound to integrate human cognition, learning, thinking, and reasoning, further enhancing its capabilities.