SGDOptimizer

  • class paddle.fluid.optimizer.SGDOptimizer(learning_rate, regularization=None, name=None)[源代码]

该接口实现随机梯度下降算法的优化器

SGDOptimizer - 图1

  • 参数:
    • learning_rate (float|Variable) - 用于更新参数的学习率。可以是浮点值,也可以是具有一个浮点值作为数据元素的变量。
    • regularization - 一个正则化器,例如 fluid.regularizer.L2DecayRegularizer
    • name (str, 可选) - 可选的名称前缀,一般无需设置,默认值为None。

代码示例

  1. import paddle
  2. import paddle.fluid as fluid
  3. import numpy as np
  4.  
  5. place = fluid.CPUPlace()
  6. main = fluid.Program()
  7. with fluid.program_guard(main):
  8. x = fluid.layers.data(name='x', shape=[13], dtype='float32')
  9. y = fluid.layers.data(name='y', shape=[1], dtype='float32')
  10. y_predict = fluid.layers.fc(input=x, size=1, act=None)
  11. cost = fluid.layers.square_error_cost(input=y_predict, label=y)
  12. avg_cost = fluid.layers.mean(cost)
  13.  
  14. sgd_optimizer = fluid.optimizer.SGD(learning_rate=0.001)
  15. sgd_optimizer.minimize(avg_cost)
  16.  
  17. fetch_list = [avg_cost]
  18. train_reader = paddle.batch(
  19. paddle.dataset.uci_housing.train(), batch_size=1)
  20. feeder = fluid.DataFeeder(place=place, feed_list=[x, y])
  21. exe = fluid.Executor(place)
  22. exe.run(fluid.default_startup_program())
  23. for data in train_reader():
  24. exe.run(main, feed=feeder.feed(data), fetch_list=fetch_list)
  • minimize(loss, startup_program=None, parameter_list=None, no_grad_set=None, grad_clip=None)

为网络添加反向计算过程,并根据反向计算所得的梯度,更新parameter_list中的Parameters,最小化网络损失值loss。

  • 参数:
    • loss (Variable) – 需要最小化的损失值变量
    • startup_program (Program, 可选) – 用于初始化parameter_list中参数的 Program , 默认值为None,此时将使用 default_startup_program
    • parameter_list (list, 可选) – 待更新的Parameter组成的列表, 默认值为None,此时将更新所有的Parameter
    • no_grad_set (set, 可选) – 不需要更新的Parameter的集合,默认值为None
    • grad_clip (GradClipBase, 可选) – 梯度裁剪的策略,静态图模式不需要使用本参数,当前本参数只支持在dygraph模式下的梯度裁剪,未来本参数可能会调整,默认值为None

返回: (optimize_ops, params_grads),数据类型为(list, list),其中optimize_ops是minimize接口为网络添加的OP列表,params_grads是一个由(param, grad)变量对组成的列表,param是Parameter,grad是该Parameter对应的梯度值

返回类型: tuple

代码示例

  1. import paddle
  2. import paddle.fluid as fluid
  3. import numpy as np
  4.  
  5. place = fluid.CPUPlace()
  6. main = fluid.Program()
  7. with fluid.program_guard(main):
  8. x = fluid.layers.data(name='x', shape=[13], dtype='float32')
  9. y = fluid.layers.data(name='y', shape=[1], dtype='float32')
  10. y_predict = fluid.layers.fc(input=x, size=1, act=None)
  11. cost = fluid.layers.square_error_cost(input=y_predict, label=y)
  12. avg_cost = fluid.layers.mean(cost)
  13.  
  14. sgd_optimizer = fluid.optimizer.SGD(learning_rate=0.001)
  15. sgd_optimizer.minimize(avg_cost)
  16.  
  17. fetch_list = [avg_cost]
  18. train_reader = paddle.batch(
  19. paddle.dataset.uci_housing.train(), batch_size=1)
  20. feeder = fluid.DataFeeder(place=place, feed_list=[x, y])
  21. exe = fluid.Executor(place)
  22. exe.run(fluid.default_startup_program())
  23. for data in train_reader():
  24. exe.run(main, feed=feeder.feed(data), fetch_list=fetch_list)