plot_clr_history: Plot cyclic learning rate callback information

In bradleyboehmke/clr: Cyclic Learning Rate

Description

Plot cyclic learning rate callback information

Usage

plot_clr_history(callback_clr, granularity = "epoch",
  trans_y_axis = "identity")

Arguments

callback_clr	An object of class 'CyclicLR'.
granularity	Either "epoch" or "iteration". We advise to use epoch as we find it easier to work with. The plot will look very similar (except for the x-axis scaling) for both options as long as you choosed step_size in new_callback_cyclical_learning_rate() to be more iterations than one epoch has.
trans_y_axis	Value passed to [ggplot2::scale_y_continuous()] as the 'trans' argument. Only supported for 'backend = "ggplot2"'.
backend	Either "base" for base R or "ggplot2".

Examples

library(keras)
dataset <- dataset_boston_housing()
c(c(train_data, train_targets), c(test_data, test_targets)) %<-% dataset

mean <- apply(train_data, 2, mean)
std <- apply(train_data, 2, sd)
train_data <- scale(train_data, center = mean, scale = std)
test_data <- scale(test_data, center = mean, scale = std)


model <- keras_model_sequential() %>%
  layer_dense(
    units = 64, activation = "relu",
    input_shape = dim(train_data)[[2]]
  ) %>%
  layer_dense(units = 64, activation = "relu") %>%
  layer_dense(units = 1)
model %>% compile(
  optimizer = optimizer_rmsprop(lr = 0.001),
  loss = "mse",
  metrics = c("mae")
)

callback_clr <- new_callback_cyclical_learning_rate(
  step_size = 32,
  base_lr = 0.001,
  max_lr = 0.006,
  gamma = 0.99,
  mode = "exp_range"
)
model %>% fit(
  train_data, train_targets,
  validation_data = list(test_data, test_targets),
  epochs = 10, verbose = 1,
  callbacks = list(callback_clr)
)
callback_clr$history
plot_clr_history(callback_clr, backend = "base")

bradleyboehmke/clr documentation built on Jan. 16, 2020, 12:49 a.m.