Papers of the day   All papers

WeatherBench: A benchmark dataset for data-driven weather forecasting


Stephan Rasp: Can AI predict weather? Let's find out! Announcing our brand-new benchmark dataset for data-driven weather forecasting: WeatherBench. We prepared the best available data and posed a clear challenge. Data and code: Paper: Thread.

9 replies, 659 likes

Bojan Tunguz: I have been urging @kaggle to setup this kind of competition for a while.

3 replies, 31 likes

Brandon Rohrer: If you’re new to time series prediction and would like to get a gut-level feel for the feature selection and modeling issues, @E2eMl Course 212 shows how to build a simple weather predictor.

1 replies, 18 likes

Stephan Hoyer: @chrmanning For a sobering sssessment of current SOTA, I recommend this pre-print on the “WeatherBench” dataset by @raspstephan and colleagues:

2 replies, 13 likes

Machine Learning Tokyo: How good are you at predicting weather? 🧙‍♂️

0 replies, 10 likes

Stephan Rasp: The leaderboard is now live on the Github repo: Here is a summary of the WeatherBench challenge 👇

0 replies, 5 likes

J. Emmanuel Johnson: Super neat! I like benchmark datasets. Otherwise it becomes really hard to gauge the effectiveness of the algorithms people are using. Plus, we get something new to play with other than digits and cats.

0 replies, 3 likes


Found on Feb 17 2020 at

PDF content of a computer science paper: WeatherBench: A benchmark dataset for data-driven weather forecasting