This post is an invite to, collectively, produce the best possible architecture and tuning for the sunspot activity and share these insights.
In week 4 of the course, we examined how to predict sunspot activity putting everything learned in this course together. Predicting sunspot activity is challenging due to its distinct long-range seasonality and noise intensity. Developing a good prediction model is hence not easy.
Happily, the sunspot activity is an open lab - so we can exchange code freely.
Call for action: Post a summary* of your model architecture and hyperparameters together with your mae on the validation set.
(* or just copy&paste the code from your lab 
)
Make a new post with your model, if its below the current best.
The best model and its person is updated regularly in this post.
Let’s all learn from each other! 
*** The current best mae: 14.00 on the validation set ***
by: Laurence
