uncertainty-lens / evaluation.json
ARotting's picture
Publish Five-member calibrated tiny vision ensemble
7671040 verified
Raw
History Blame Contribute Delete
1.92 kB
{
"model": "Uncertainty Lens Deep Ensemble",
"members": 5,
"parameters_per_member": 2198,
"total_parameters": 10990,
"member_training": [
{
"member": 0,
"seed": 2036,
"best_validation_accuracy": 0.9814814925193787,
"best_epoch": 29
},
{
"member": 1,
"seed": 2037,
"best_validation_accuracy": 0.9777777791023254,
"best_epoch": 22
},
{
"member": 2,
"seed": 2038,
"best_validation_accuracy": 0.9740740656852722,
"best_epoch": 50
},
{
"member": 3,
"seed": 2039,
"best_validation_accuracy": 0.9777777791023254,
"best_epoch": 30
},
{
"member": 4,
"seed": 2040,
"best_validation_accuracy": 0.9777777791023254,
"best_epoch": 42
}
],
"temperature": 1.0,
"single_member_test": {
"accuracy": 0.9814814814814815,
"negative_log_likelihood": 0.05506124719977379,
"brier_score": 0.027842911624812734,
"expected_calibration_error": 0.021852720225298845
},
"ensemble_uncalibrated_test": {
"accuracy": 0.9925925925925926,
"negative_log_likelihood": 0.05701135843992233,
"brier_score": 0.024981458112583597,
"expected_calibration_error": 0.025468972656461943
},
"ensemble_calibrated_test": {
"accuracy": 0.9925925925925926,
"negative_log_likelihood": 0.05701135843992233,
"brier_score": 0.024981458112583597,
"expected_calibration_error": 0.025468972656461943
},
"ood_detection": {
"composition": {
"uniform_noise": 270,
"pixel_scrambled": 270
},
"entropy_roc_auc": 0.9771399176954731,
"clean_mean_predictive_entropy": 0.10513244569301605,
"ood_mean_predictive_entropy": 1.1125800609588623,
"clean_mean_mutual_information": 0.01804155483841896,
"ood_mean_mutual_information": 0.4621701240539551
}
}