阿里天池—2022江苏气象预测AI算法挑战赛

原创

已于 2022-05-08 20:51:50 修改 · 3k 阅读

12 ·

CC 4.0 BY-SA版权

文章标签：

#人工智能 #算法 #transformer

于 2022-05-08 18:50:55 首次发布

本文介绍了一个利用MAE预训练模型和Transformer进行气象预测的比赛策略。通过获取三种不同的图像编码，结合优化后的Transformer进行时序预测，以图像识别和处理为目标。数据集包含雷达、降水和平均风的图像，模型需要基于过去20张图片预测未来20张。MAE用于提取图像信息，Transformer则实现时间序列预测。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

文章目录

摘要
一、数据分析
二、MAE简介
三、Transformer简介
四、模型搭建(还未写......)

摘要

This is a meteorological forecasting competition being held by Ali Tianchi. The contestants need to predict future meteorological data based on past meteorological data; the meteorological data is given in the form of pictures, including radar echo map, rainfall map, average wind image, so it is necessary to use the technology of image information extraction; the initial idea is to obtain three different Encodings and Decoders through pre-training through the MAE model. These three Encodings will understand the meaning of the image data they are responsible for and keep the three Encodings The parameters remain unchanged, and an optimized Transformer model is combined behind them. The Transformer model will realize the function of completing time series prediction, and the predicted data will be finally decoded into the meteorological data in the picture format through the Decoder in MAE, so as to achieve image recognition and processi