Hi, I have a question about the TNG100-1 dataset, specifically regarding the (r) Galaxy Morphologies (Deep Learning) section in the data specifications. I understand that in Huertas-Company et al. (2019), galaxies were classified into four morphological types: E (Elliptical), S0/a (Lenticular), Sab (Early-Type Spiral), and Scd (Late-Type Spiral). However, in the actual HDF5 files, only the probabilities for P_Late, P_S0, and P_Sab are provided. Could you clarify why the probabilities for E (Elliptical) and Scd (Late-Type Spiral) are not included in the dataset?
The probabilities provided in the catalo are the outputs of binary classifiers.
To compute the probabilities of the different subclasses used in Huertas-Company et al. (2019), you can use the following relations - If I recall properly.
Hi, I have a question about the TNG100-1 dataset, specifically regarding the (r) Galaxy Morphologies (Deep Learning) section in the data specifications. I understand that in Huertas-Company et al. (2019), galaxies were classified into four morphological types: E (Elliptical), S0/a (Lenticular), Sab (Early-Type Spiral), and Scd (Late-Type Spiral). However, in the actual HDF5 files, only the probabilities for P_Late, P_S0, and P_Sab are provided. Could you clarify why the probabilities for E (Elliptical) and Scd (Late-Type Spiral) are not included in the dataset?
Hello,
The probabilities provided in the catalo are the outputs of binary classifiers.
To compute the probabilities of the different subclasses used in Huertas-Company et al. (2019), you can use the following relations - If I recall properly.
Best,
Marc