Models will be evaluated on their ability to accurately predict the spatial cell-type composition across the test tissue slice. Each model will be scored by the [**Spearman rank correlation**](https://en.wikipedia.org/wiki/Spearman%27s_rank_correlation_coefficient) between the predicted and ground truth cell type abundance vectors. The Spearman score will be calculated for each test spot first, then averaged across spots to obtain an aggregated score. 

The test slide has been divided into a *public test set* and a *private test set*. The aggregated score will be computed on the two groups separately. The cellular profiles of these two groups are weakly correlated, so performance on the public set would not necessarily reflect the projected performance on the private test set. The identity of each test spot (i.e. whether it falls in the public or private set) is provided along with the spot 2D coordinates in the Challenge data .h5 file.

To rank submissions on the final leaderboard, the aggregated Spearman score on the held-out *private test set* will be used. This leaderboard will be released only upon the end of the Challenge (post the final submission deadline). 

While the challenge is ongoing, teams will have an opportunity to assess where they stand on a public leaderboard, which will be based on scores on the *public test set*. The submission file must contain the predictions on both the public and private sets of spots, with the spot ids expected in the same order as in the Test slide (slide id S_7). Team scores on the private test set (which will be used for assigning the final team ranks) will only be revealed after the end date of the Challenge.


## Submission Requirements
### This is a Code Competition
Submissions to this competition must be made through Notebooks. In order for the "Submit" button to be active after a commit, the following conditions must be met:

CPU Notebook <= 9 hours run-time

GPU Notebook <= 9 hours run-time

Internet access disabled

Submission file must be named *submission.csv*. Predictions for each test sample (spot) are expected to be provided as a comma-separated list of floats (predictions for one spot per line) in the following format: `ID,C1,C2,… C35`, where ID is the identifier for each spot in the test set, and C1-C35 are the abundances of the 35 possible cell types assigned to the spot.