Early Termination¶
Callbacks and event handlers used to terminate training as soon as the running loss becomes lower than the theoretical-smallest.
Tensorflow v2¶
Tensorflow v2 code to terminate training of a deep learning regressor or classifier when the running loss is much lower than a threshold, typically the theoretical-best.
- class kxy.learning.tensorflow_early_termination.TerminateIfOverfittedTF(theoretical_best, loss_key)¶
Tensorflow callback that terminates training at the end of a batch when the running loss is smaller than the theoretical best, which is strong indication that the model will end up overfitting.
- Parameters
loss_key (str) – Which loss to base early-termination on. Example values are:
'loss','classification_error', and any other registered loss metrics.theoretical_best (float) – The theoretical-smallest loss achievable without overfiting, obtained using
df.kxy.data_valuation
- on_batch_end(batch, logs=None)¶
PyTorch¶
PyTorch code to terminate training of a deep learning regressor or classifier when the running loss is much lower than a threshold, typically the theoretical-best.
- class kxy.learning.pytorch_early_termination.TerminateIfOverfittedPT(theoretical_best, loss_key)¶
PyTorch event handler that terminates training when the running loss is smaller than the theoretical best, which is strong indication that the model will end up overfitting.
- Parameters
loss_key (str) – Which loss to base early-termination on. Example values are:
'loss','classification_error', and any other registered loss metrics.theoretical_best (float) – The theoretical-smallest loss achievable without overfiting, obtained using
df.kxy.data_valuation.