Rishi Gupta
@rishigupta.bsky.social
Respiratory physician, NIHR advanced fellow, UCL-TB, UCL Respiratory
2. Proprietary models eg Epic deterioration index. If the company are not willing/able to make the model reproducible, do we still want to see validation data?
December 11, 2024 at 2:44 PM
2. Proprietary models eg Epic deterioration index. If the company are not willing/able to make the model reproducible, do we still want to see validation data?
I agree. 2 problematic areas on this:
1. Machine learning models where coefficients cannot readily be derived. Should we insist that the model object is shared in these cases? Does this create issues if the object includes the training data (can be tricky to fully remove)?
1. Machine learning models where coefficients cannot readily be derived. Should we insist that the model object is shared in these cases? Does this create issues if the object includes the training data (can be tricky to fully remove)?
December 11, 2024 at 2:43 PM
I agree. 2 problematic areas on this:
1. Machine learning models where coefficients cannot readily be derived. Should we insist that the model object is shared in these cases? Does this create issues if the object includes the training data (can be tricky to fully remove)?
1. Machine learning models where coefficients cannot readily be derived. Should we insist that the model object is shared in these cases? Does this create issues if the object includes the training data (can be tricky to fully remove)?