Ftrl
classkeras.optimizers.Ftrl(
learning_rate=0.001,
learning_rate_power=-0.5,
initial_accumulator_value=0.1,
l1_regularization_strength=0.0,
l2_regularization_strength=0.0,
l2_shrinkage_regularization_strength=0.0,
beta=0.0,
weight_decay=None,
clipnorm=None,
clipvalue=None,
global_clipnorm=None,
use_ema=False,
ema_momentum=0.99,
ema_overwrite_frequency=None,
loss_scale_factor=None,
gradient_accumulation_steps=None,
name="ftrl",
**kwargs
)
优化器实现了FTRL算法.
"Follow The Regularized Leader”(FTRL)是Google在2010年代初期为点击率预测开发的一种优化算法.它最适合用于具有大而稀疏特征空间的浅层模型.该算法由McMahan et al., 2013描述.Keras版本支持在线L2正则化(论文中描述的L2正则化)和收缩型L2正则化(这是对损失函数添加L2惩罚).
初始化:
n = 0
sigma = 0
z = 0
单个变量w
的更新规则:
prev_n = n
n = n + g ** 2
sigma = (n ** -lr_power - prev_n ** -lr_power) / lr
z = z + g - sigma * w
if abs(z) < lambda_1:
w = 0
else:
w = (sgn(z) * lambda_1 - z) / ((beta + sqrt(n)) / alpha + lambda_2)
符号说明:
lr
是学习率g
是变量的梯度lambda_1
是L1正则化强度lambda_2
是L2正则化强度lr_power
是缩放n的幂.有关启用收缩时的更多详细信息,请查看l2_shrinkage_regularization_strength
参数的文档,在这种情况下,梯度被替换为带有收缩的梯度.
参数:
learning_rate: 一个浮点数,一个
keras.optimizers.schedules.LearningRateSchedule
实例,或
一个不带参数并返回实际值的调用.学习率.默认为0.001
.
learning_rate_power: 一个浮点数值,必须小于或等于零.
控制学习率在训练期间的下降.使用零表示固定学习率.
initial_accumulator_value: 累加器的起始值.只允许零或正数值.
l1_regularization_strength: 一个浮点数值,必须大于或等于零.默认为0.0
.
l2_regularization_strength: 一个浮点数值,必须大于或等于零.默认为0.0
.
l2_shrinkage_regularization_strength: 一个浮点数值,必须大于或等于零.这与上面的L2不同,因为上面的L2是一种稳定化惩罚,而这种L2收缩是一种幅度惩罚.当输入稀疏时,收缩只会发生在活动权重上.
beta: 一个浮点数值,表示论文中的beta值.默认为0.0
.
name: String. The name to use
for momentum accumulator weights created by
the optimizer.
weight_decay: Float. If set, weight decay is applied.
clipnorm: Float. If set, the gradient of each weight is individually
clipped so that its norm is no higher than this value.
clipvalue: Float. If set, the gradient of each weight is clipped to be
no higher than this value.
global_clipnorm: Float. If set, the gradient of all weights is clipped
so that their global norm is no higher than this value.
use_ema: Boolean, defaults to False
.
If True
, exponential moving average
(EMA) is applied. EMA consists of computing an exponential moving
average of the weights of the model (as the weight values change
after each training batch), and periodically overwriting the
weights with their moving average.
ema_momentum: Float, defaults to 0.99. Only used if use_ema=True
.
This is the momentum to use when computing
the EMA of the model's weights:
new_average = ema_momentum * old_average + (1 - ema_momentum) *
current_variable_value
.
ema_overwrite_frequency: Int or None, defaults to None. Only used if
use_ema=True
. Every ema_overwrite_frequency
steps of iterations,
we overwrite the model variable by its moving average.
If None, the optimizer
does not overwrite model variables in the middle of training,
and you need to explicitly overwrite the variables
at the end of training by calling
optimizer.finalize_variable_values()
(which updates the model
variables in-place). When using the built-in fit()
training loop,
this happens automatically after the last epoch,
and you don't need to do anything.
loss_scale_factor: Float or None
. If a float, the scale factor will
be multiplied the loss before computing gradients, and the inverse
of the scale factor will be multiplied by the gradients before
updating variables. Useful for preventing underflow during
mixed precision training. Alternately,
keras.optimizers.LossScaleOptimizer
will
automatically set a loss scale factor.
gradient_accumulation_steps: Int or None
. If an int, model & optimizer
variables will not be updated at every step; instead they will be
updated every gradient_accumulation_steps
steps, using the average
value of the gradients since the last update. This is known as
"gradient accumulation". This can be useful
when your batch size is very small, in order to reduce gradient
noise at each update step. EMA frequency will look at "accumulated"
iterations value (optimizer steps // gradient_accumulation_steps).
Learning rate schedules will look at "real" iterations value
(optimizer steps).