Prediction of logarithmic Henry coefficient for carbon dioxide (out-of-distribution)#
Task description#
The Henry coefficient is the slope of the gas adsorption isotherm for \(p\to0\). It can be computed using grand canonical Monte Carlo. For this task, we use the data computed in [Moosavi2021] where the framework is modeled using the UFF force field [UFF] and the carbon dioxide using the TraPPE [Trappe] force field.
To estimate the predictive performance of models on out-of-distribution data, this task uses a splitter that splits the data based on quantile bins of the density.
Overview#
Leaderboard#
ID |
Name |
Mean Squared Error |
Mean Absolute Error |
R2 Score |
Max Error |
Top 50 In Top 50 |
Top 100 In Top 100 |
---|---|---|---|---|---|---|---|
default-xgboost-default-feat |
0.95 |
0.75 |
-0.0 |
3.58 |
0 |
0 |
|
dummy-median-model |
1.14 |
0.84 |
-0.18 |
3.19 |
0 |
0 |
|
dummy-mean-model |
1.14 |
0.84 |
-0.18 |
3.19 |
0 |
0 |
|
dummy-mean-model |
1.14 |
0.84 |
-0.18 |
3.19 |
0 |
0 |