Weights stability comparison.
- Michela Carlotta Massi (9122257)
- Lorenzo Dominoni (13864268)
- Francesca Ieva (144652)
- Giovanni Fiorito (136942)
Publication date
September 2022
DOI Publisher
Public Library of Science (PLoS)Abstract
Importance weights stability performance (KT-stability) of all the tested Deep Survival NN architectures (blue) and all multivariate CoxPH models fitted. The values in parenthesis for Deep Survival NNs represent the number of input nodes (i.e. the granularity of features’ clusters) and the number of nodes in the last layer before the output. Whereas the parenthesis for CoxPH models report the granularity of the input features’ clusters. Dots represent the average performance value, while bands report the confidence intervals around the mean computed on the K = 10 splits. (PNG)</p
Add to ORKG