tf.keras.experimental.NoisyLinearCosineDecay

A LearningRateSchedule that uses a noisy linear cosine decay schedule.

Inherits From: LearningRateSchedule

View aliases

Compat aliases for migration

See Migration guide for more details.

tf.compat.v1.keras.experimental.NoisyLinearCosineDecay

tf.keras.experimental.NoisyLinearCosineDecay(
    initial_learning_rate, decay_steps, initial_variance=1.0, variance_decay=0.55,
    num_periods=0.5, alpha=0.0, beta=0.001, name=None
)

See [Bello et al., ICML2017] Neural Optimizer Search with RL. https://arxiv.org/abs/1709.07417

For the idea of warm starts here controlled by num_periods, see [Loshchilov & Hutter, ICLR2016] SGDR: Stochastic Gradient Descent with Warm Restarts. https://arxiv.org/abs/1608.03983

Note that linear cosine decay is more aggressive than cosine decay and larger initial learning rates can typically be used.

When training a model, it is often recommended to lower the learning rate as the training progresses. This schedule applies a noisy linear cosine decay function to an optimizer step, given a provided initial learning rate. It requires a step value to compute the decayed learning rate. You can just pass a TensorFlow variable that you increment at each training step.

The schedule a 1-arg callable that produces a decayed learning rate when passed the current optimizer step. This can be useful for changing the learning rate value across different invocations of optimizer functions. It is computed as:

def decayed_learning_rate(step):
  step = min(step, decay_steps)
  linear_decay = (decay_steps - step) / decay_steps)
  cosine_decay = 0.5 * (
      1 + cos(pi * 2 * num_periods * step / decay_steps))
  decayed = (alpha + linear_decay + eps_t) * cosine_decay + beta
  return initial_learning_rate * decayed

where eps_t is 0-centered gaussian noise with variance initial_variance / (1 + global_step) ** variance_decay

Example usage:

decay_steps = 1000
lr_decayed_fn = (
  tf.keras.experimental.NoisyLinearCosineDecay(
    initial_learning_rate, decay_steps))

You can pass this schedule directly into a tf.keras.optimizers.Optimizer as the learning rate. The learning rate schedule is also serializable and deserializable using tf.keras.optimizers.schedules.serialize and tf.keras.optimizers.schedules.deserialize.

Methods

from_config

View source

@classmethod
from_config(
    config
)

Instantiates a LearningRateSchedule from its config.

get_config

View source

get_config()

__call__

View source

__call__(
    step
)

Call self as a function.