Genomes to Fields Prediction Competition

Objective: Develop models to predict corn yield in the 2024 G2F trials based on the current G2F dataset and other publicly available data

About

I found out about this competition recently (Nov 2024) while on LinkedIn. Competitions are a good way to test ones skills.

Interestingly, the scoring criteria is a weighted Pearson Correlation Coefficient (\(\rho\)). I think a primary challenge will be to ensure the modeling process accounts for this scoring metric. For example, a classic regression loss function is RMSE but I’m unsure how well \(\rho\) relates to RMSE. I’ve never tried to train a model using \(\rho\). I know it’s possible to use \(R^2\) as a loss function. Perhaps that’s what I’ll try. To be updated.

The rules also state that any public data can be used as long as it was available before Feb 1, 2024. So part of the challenge will be to assess what external data might be useful in addtion to the data the competition provides.

Finally, there’s a $4k reward 😎