-
|
Hi CEBRA team! First of all, thanks for creating such a powerful and user-friendly tool. I'm using the consistency scores to see how stable the Cebra-behavior models trained on different animals are. I have seen some very different scores in the same pair of animals. I noticed the different scores in the demo as well, but they were less dramatic than mine. I'm assuming the difference came from the two labels for each animal, but the documentation suggests that "all sets of digitalized labels contain the same set of values". I would appreciate if more information can be shared on how the consistency scores are calculated. Thanks! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
Hi @Owenxz , for a list of embeddings X1, ..., Xn each entry in the matrix is the value of regressing from Xi -> Xj, and reporting the R2 value of that linear regression. This R2 can be different for the Xi->Xj and Xj->Xi regression problems. If you are not interested in considering this difference, you could also compute the average between the upper and lower parts. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you so much! That makes a lot of sense. |
Beta Was this translation helpful? Give feedback.
Hi @Owenxz , for a list of embeddings X1, ..., Xn each entry in the matrix is the value of regressing from Xi -> Xj, and reporting the R2 value of that linear regression.
This R2 can be different for the Xi->Xj and Xj->Xi regression problems. If you are not interested in considering this difference, you could also compute the average between the upper and lower parts.