will it need more epochs for training or it is not a necessary and what is the learning rate I should set for this data with optimizer adam?
1 Answer
$\begingroup$
$\endgroup$
3
you can use the default learning rate, then apply the learning rate scheduler to adjust your learning rate in every few epochs. Please refer to the keras document
-
$\begingroup$ This is alternative approach. You can try one of these. You can use the adam default learning rate in kears as the initial_learning_rate then it will adjust/decay in every few epochs/steps. $\endgroup$– DaCardCommented Feb 23, 2022 at 18:03
-
$\begingroup$ thanks for replying but excuse what about decay_steps and decay_rate? and does the epochs will exceed 20 epochs as a dataset is large $\endgroup$– samCommented Feb 23, 2022 at 18:13
-
$\begingroup$ These are all hyper-parameters. You need to specify in advance. I encourage you to try out few different numbers. (# of steps) * (batch_size) = 1 epoch, it all depends on your data size. Try several sets of decay_steps, decay_rates. Again these are hyper-parameters. BTW, if you think my solution is helpful, please accept my answer and hit up. Thanks $\endgroup$– DaCardCommented Feb 23, 2022 at 18:37