tuneCLR: Tuning the cyclical learning rate (CLR)

In kristianHoden/smartPARE: An R package for efficient identification of true mRNA cleavage sites

Description

The function is run to find the minimum and maximum values of the cyclical learning rate

Usage

tuneCLR(
  batch_size2 = 64,
  epochs_find_LR = 20,
  lr_max = 0.1,
  optimizer2 = keras::optimizer_rmsprop(lr = lr_max, decay = 0),
  validation_split2 = 0.2,
  rollmeanSplit = 3
)

Arguments

batch_size2	Batch size
epochs_find_LR	Epochs to run to find the optimal learnig rate, often a low number is needed
lr_max	Maximum approved learning rate
optimizer2	Optimizer
validation_split2	Fraction of the data used to validate the model
rollmeanSplit	smothens the curve by rollmean division of this number

Examples

tuneCLR(batch_size2 = 64,
epochs_find_LR = 20,
lr_max = 0.1,
optimizer2 = optimizer_rmsprop(lr=lr_max, decay=0),
validation_split2 =0.2,
rollmeanSplit = 3
)

kristianHoden/smartPARE documentation built on July 3, 2021, 7:10 p.m.