Python 千层面的学习率衰减

我使用nesterov动量来更新卷积神经网络的权重。我在用千层面做CNN。如何实现每个时代的学习率衰减

prediction = lasagne.layers.get_output(network)
loss = lasagne.objectives.categorical_crossentropy(prediction, target_var)
params = lasagne.layers.get_all_params(network, trainable=True)
lasagne.updates.nesterov_momentum(loss, params, learning_rate, momentum=0.9)
train_fn = theano.function([input_var_1, input_var_2, target_var], loss, updates=updates)

---

我想你遗漏了什么：

prediction = lasagne.layers.get_output(network)
loss = lasagne.objectives.categorical_crossentropy(prediction, target_var)
params = lasagne.layers.get_all_params(network, trainable=True)
updates = lasagne.updates.nesterov_momentum(loss, params, learning_rate, momentum=0.9)
train_fn = theano.function([input_var_1, input_var_2, target_var], loss, updates=updates)

---

我想你遗漏了什么：

prediction = lasagne.layers.get_output(network)
loss = lasagne.objectives.categorical_crossentropy(prediction, target_var)
params = lasagne.layers.get_all_params(network, trainable=True)
updates = lasagne.updates.nesterov_momentum(loss, params, learning_rate, momentum=0.9)
train_fn = theano.function([input_var_1, input_var_2, target_var], loss, updates=updates)

---

将您的学习率定义为图形元素

...
learning_rate = T.scalar(name='learning_rate')
updates = lasagne.updates.nesterov_momentum(loss, params, learning_rate, momentum=0.9)
...

将您的学习率节点添加到培训中

train_fn = theano.function([input_var_1, input_var_2, target_var, learning_rate], loss, updates=updates)

现在，您可以调整学习速率，例如：

base_lr = 1e-4
lr_decay = 0.8

for epoch in range(epochs):
    lr = base_lr * (lr_decay ** epoch)
    loss, updates = train_fn(input_1, input_2, target, lr)

---

将您的学习率定义为图形元素

...
learning_rate = T.scalar(name='learning_rate')
updates = lasagne.updates.nesterov_momentum(loss, params, learning_rate, momentum=0.9)
...

将您的学习率节点添加到培训中

train_fn = theano.function([input_var_1, input_var_2, target_var, learning_rate], loss, updates=updates)

现在，您可以调整学习速率，例如：

base_lr = 1e-4
lr_decay = 0.8

for epoch in range(epochs):
    lr = base_lr * (lr_decay ** epoch)
    loss, updates = train_fn(input_1, input_2, target, lr)

---

你能像Keras那样使用adam吗？如果是这样的话，你可以使用adam或one，如果它的变体…你能像Keras一样使用adam吗？如果是这样，你可以使用adam或one，如果它的变体。。。