Not finished
TBD
We offer two ways to evaluate the performance of model.
- Tensorboard, Tensorboard records the accuracy, specificity, precision, recall, f1, and time deviation of training set (mini batch) and test set respectively.
- CSV. Csv file records the accuracy, specificity, precision, recall, f1, hamming loss, coverage, ranking loss, average precision, macro auc, micro auc, time deviation of training set (mini batch) and test set respectively
Launch tensorboard: type 'tensorboard --logdir={root path}\model_evaluate{time(millisecond)}' in cmd. Once TensorBoard is running, navigate web browser to localhost:6006 to view the result. more details in link
Csv. Training result is saved in {root path}\model_evaluate{time(millisecond)}\train while the test result is saved in {root path}\model_evaluate{time(millisecond)}\test.