|
pulsatrix 1.1.0
|
A fitted temperature and the weighted mean log loss before and after it. More...
#include <calibration.hpp>
Public Attributes | |
| double | temperature = 1.0 |
| double | log_loss_before = 0.0 |
| at T = 1 | |
| double | log_loss_after = 0.0 |
| at the fitted T | |
| int64_t | iterations = 0 |
| bool | converged = true |
| False when the minimum is at a bound of T in [1e-6, 1e6]: the loss keeps falling as T grows (the logits are no better than chance, or point the wrong way) or as T shrinks (the logits separate the labels perfectly). temperature is then that bound. | |
A fitted temperature and the weighted mean log loss before and after it.
| bool pulsatrix::TemperatureFit::converged = true |
False when the minimum is at a bound of T in [1e-6, 1e6]: the loss keeps falling as T grows (the logits are no better than chance, or point the wrong way) or as T shrinks (the logits separate the labels perfectly). temperature is then that bound.
| int64_t pulsatrix::TemperatureFit::iterations = 0 |
| double pulsatrix::TemperatureFit::log_loss_after = 0.0 |
at the fitted T
| double pulsatrix::TemperatureFit::log_loss_before = 0.0 |
at T = 1
| double pulsatrix::TemperatureFit::temperature = 1.0 |