You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The nature of agreement test failure is disagreement of features on which classifier is trained. This happens because simulated data is not ideal: unfortunately, we cannot take into account all processes in a detector during simulation.
Thus, to pass agreement check you can do (the simplest and obvious things):
train classifier only on 'good' features
add noise to your submission (noise.ipynb - will generate baseline_noise.csv)
(your ideas here)
Passing the agreement test is an important part of the challenge!