Principle of Machine Learning and Its Potential Application in Cli-mate Prediction

Shengping He, Huijun Wang, Hua Li, Jiazhen Zhao

Journal of Autonomous Intelligence · 11 March 2021

In one sentence. Tutorial on machine learning basics, with a CNN applied to East Asian winter monthly temperature and compared against a dynamical model; no headline numeric result.

Abstract

After two “cold winters of artificial intelligence”, machine learning has once again entered the public’s vision in recent ten years, and has a momentum of rapid development. It has achieved great success in practical applications such as image recognition and speech recognition system. It is one of the main tasks and objectives of machine learning to summarize key information and main features from known data sets, so as to accurately identify and predict new data. From this perspective, the idea of integrating machine learning into climate prediction is feasible. Firstly, taking the adjustment of linear fitting parameters (i.e. slope and intercept) as an example, this paper introduces the process of machine learning optimizing parameters through gradient descent algorithm and finally obtaining linear fitting function. Secondly, this paper introduces the construction idea of neural network and how to apply neural network to fit nonlinear function. Finally, the framework principle of convolutional neural network for deep learning is described, and the convolutional neural network is applied to the return test of monthly temperature in winter in East Asia, and compared with the return results of climate dynamic model. This paper will help to understand the basic principle of machine learning and provide some reference ideas for the application of machine learning to climate prediction.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.