Results
This page explains how to interpret the results of the submitted job
After the submitted job finishes successfully without errors, the /livetraining page will show View Results button through which you can navigate to the results page.
Training Results
The following results are available on the dashboard itself:
Average training and validation accuracy across all sites
Tensorboard metrics on validation accuracy and loss versus epochs for each site
Detailed accuracy table for training data and validation data. The columns of each table represent the model whereas the row of each table represents the client site
After you click the Download Results button, a ZIP file with the following content will be downloaded:
cross_val_results.json- This file contains all the tracked metrics by site and model. Tracked metrics include:train_accuracy - Accuracy on training data
val_accuracy - Accuracy on validation data
val_sensitivity - Sensitivity on validation data
val_specificity - Sensitivity on validation data
val_npv - Negative Predictive Value on validation data
auc - Area under the ROC curve on validation data
fpr - False Positive Rate on validation data
tpr - True Positive Rate on validation data
logs.txt- Contains the process logs of communication in the federated processFL_global_model.pt- Model weights of the final global model at the end of the training process.best_FL_global_model.pt- Model weights of the best global model in the training process. Best model is determined by the highest accuracy. If you would like to modify the definition of best, please contact us to know more.tb_events- Contains the tensorboard metrics of validation accuracy and loss against local epochs at each site.
Validation Results
The following results are available on the dashboard itself:
Average accuracy across all sites
Accuracy of model at each site
After you click the Download Results button, a ZIP file with the following content will be downloaded:
cross_val_results.json- This file contains all the tracked metrics by site. Tracked metrics include:accuracy
sensitivity
specificity
npv - Negative Predictive Value
auc - Area under the ROC curve
fpr - False Positive Rate
tpr - True Positive Rate
logs.txt- Contains the process logs of communication in the federated process
Note - If you need any other custom metrics in the process, please contact us and we will implement it as soon as possible for you.
Last updated