Python 阻止渐变在关节损失中流动
我有一个输入张量Python 阻止渐变在关节损失中流动,python,tensorflow,gradient-descent,backpropagation,loss-function,Python,Tensorflow,Gradient Descent,Backpropagation,Loss Function,我有一个输入张量 data=tf.placeholder(tf.int32,[None]) 它将被嵌入到 embedding_matrix = tf.get_variable("embedding_matrix", [5,3], tf.float32, initializer=tf.random_normal_initializer()) input_vectors = tf.nn.embedding_lookup(params=embedding_matrix, ids=data) 我使用o
data=tf.placeholder(tf.int32,[None])
它将被嵌入到
embedding_matrix = tf.get_variable("embedding_matrix", [5,3], tf.float32, initializer=tf.random_normal_initializer())
input_vectors = tf.nn.embedding_lookup(params=embedding_matrix, ids=data)
我使用output1\u权重对输入向量执行线性变换,以获得network\u output1
output1_weights = tf.get_variable("output1", [3,4], tf.float32, initializer=tf.random_normal_initializer())
network_output1 = tf.matmul(input_vectors, output1_weights)
损失将是非常标准的东西
loss1 = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=output1, logits=network_output1)
现在我想使用logits网络输出1
作为输入来计算另一个线性变换
output2_weights = tf.get_variable("output2", [4,5], tf.float32, initializer=tf.random_normal_initializer())
network_output2 = tf.matmul(network_output1, output2_weights)
第二次输出的交叉熵损失
loss2 = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=output2, logits=network_output2)
这就是我想要实现的目标。在关节损耗设置中,当最小化loss1
的损耗时,我只想支持output1\u权重的梯度,当最小化loss2
时,只支持output2\u权重的梯度。换句话说,在优化loss2
时,我不希望梯度全部流回到篡改输出1_权重。我知道optimizer类中的compute_gradients
函数可以接受一个参数var_list
,但它似乎无法阻止梯度的流动,导致单独的损失。此外,我可以考虑分离损失和尽量减少它们,这也将是一个坏的解决方案在我的设置。 您所要做的就是选择一个可训练变量,并将其分配给var\u列表
首先计算不同损失的可训练变量
import numpy as np
import tensorflow as tf
data = tf.placeholder(tf.int32, [None])
output1 = tf.placeholder(tf.int32, [None])
output2 = tf.placeholder(tf.int32, [None])
embedding_matrix = tf.get_variable("embedding_matrix", [5,3], tf.float32, initializer=tf.random_normal_initializer())
input_vectors = tf.nn.embedding_lookup(params=embedding_matrix, ids=data)
# count
params_num0 = len(tf.trainable_variables())
output1_weights = tf.get_variable("output1", [3,4], tf.float32, initializer=tf.random_normal_initializer())
network_output1 = tf.matmul(input_vectors, output1_weights)
# count
params_num1 = len(tf.trainable_variables())
loss1 = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=output1, logits=network_output1)
output2_weights = tf.get_variable("output2", [4,5], tf.float32, initializer=tf.random_normal_initializer())
network_output2 = tf.matmul(network_output1, output2_weights)
loss2 = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=output2, logits=network_output2)
然后打印它们和所有可训练变量
params = tf.trainable_variables()
print(params_num0)
print(params_num1)
print(params)
# 1
# 2
# [<tf.Variable 'embedding_matrix:0' shape=(5, 3) dtype=float32_ref>, <tf.Variable 'output1:0' shape=(3, 4) dtype=float32_ref>, <tf.Variable 'output2:0' shape=(4, 5) dtype=float32_ref>]
接下来,指定相应变量的更新梯度
opt = tf.train.AdamOptimizer(0.01)
grads_vars = opt.compute_gradients(loss1,var_list=params1)
grads_vars2 = opt.compute_gradients(loss2,var_list=params2)
print(grads_vars)
print(grads_vars2)
# [(<tf.Tensor 'gradients/MatMul_grad/tuple/control_dependency_1:0' shape=(3, 4) dtype=float32>, <tf.Variable 'output1:0' shape=(3, 4) dtype=float32_ref>)]
# [(<tf.Tensor 'gradients_1/MatMul_1_grad/tuple/control_dependency_1:0' shape=(4, 5) dtype=float32>, <tf.Variable 'output2:0' shape=(4, 5) dtype=float32_ref>)]
train_op = opt.apply_gradients(grads_vars+grads_vars2)
实验
data_np = np.random.normal(size=(100))
output1_np = np.random.randint(0,4,size=(100))
output2_np = np.random.randint(0,5,size=(100))
feed_dict_v = {data: data_np, output1: output1_np, output2: output2_np}
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for i in range(2):
print("epoch:{}".format(i))
sess.run(train_op, feed_dict=feed_dict_v)
print("embedding_matrix value:\n",sess.run(embedding_matrix, feed_dict=feed_dict_v))
print("output1_weights value:\n",sess.run(output1_weights, feed_dict=feed_dict_v))
print("output2_weights value:\n",sess.run(output2_weights, feed_dict=feed_dict_v))
结果是:
epoch:0
embedding_matrix value:
[[ 0.7646786 -0.44221798 -1.6374763 ]
[-0.4061512 -0.70626575 0.09637168]
[ 1.3499098 0.38479885 -0.10424987]
[-1.3999717 0.67008936 1.8843309 ]
[-0.11357951 -1.1893668 1.1205566 ]]
output1_weights value:
[[-0.22709225 0.70598644 0.10429419 -2.2737694 ]
[-0.6364337 -0.08602498 1.9750406 0.8664075 ]
[ 0.3656631 -0.25182125 -0.14689662 -0.03764082]]
output2_weights value:
[[ 0.00554644 -0.49370843 -0.75148153 0.6645286 1.0131303 ]
[ 0.21612553 0.07851358 0.05937392 -0.3236267 -0.8081816 ]
[ 0.82237226 0.17242427 -1.3059226 -1.1134574 0.22402465]
[-1.6996336 -0.58993673 -0.7071007 0.8407903 0.62416744]]
epoch:1
embedding_matrix value:
[[ 0.7646786 -0.44221798 -1.6374763 ]
[-0.4061512 -0.70626575 0.09637168]
[ 1.3499098 0.38479885 -0.10424987]
[-1.3999717 0.67008936 1.8843309 ]
[-0.11357951 -1.1893668 1.1205566 ]]
output1_weights value:
[[-0.21710345 0.6959941 0.11408082 -2.2637703 ]
[-0.64639646 -0.07603455 1.9650643 0.85640883]
[ 0.35567763 -0.24182947 -0.15682784 -0.04763966]]
output2_weights value:
[[ 0.01553426 -0.5036415 -0.7415529 0.65454334 1.003145 ]
[ 0.20613036 0.08847766 0.04942677 -0.31363514 -0.7981894 ]
[ 0.8323502 0.16245098 -1.2959852 -1.1234138 0.21408063]
[-1.6896346 -0.59990865 -0.6971453 0.8307945 0.6141711 ]]
您可以看到嵌入矩阵
从未更改。输出1\u权重
和输出2\u权重
仅更新相应的渐变
添加
实际上,您可以在output2\u权重上组合loss1
和loss2
。例如:
grads_vars3 = opt.compute_gradients(loss1+loss2,var_list=params2)
当通过加法组合loss1
和loss2
时,您会发现grads\u vars2
和grads\u vars3
是相等的。原因是loss1
的梯度不会流向loss1+loss2
中的output2\u权重。但在以下情况下,grads\u vars2
和grads\u vars3
在通过乘法组合loss1
和loss2
时不相等
grads_vars3 = opt.compute_gradients(loss1*loss2,var_list=params2)
上述情况意味着我们可以根据自己的需要将相应可训练变量的损失合并起来
在您的场景中,网络输出2
需要使用网络输出1
,因此我们必须指定损耗。如果网络输出2
不依赖于网络输出1
,我们可以直接优化loss1+loss2
关于渐变
input = tf.constant([[1,2,3]],tf.float32)
label1 = tf.constant([[1,2,3,4]],tf.float32)
label2 = tf.constant([[1,2,3,4,5]],tf.float32)
weight1 = tf.reshape(tf.range(12,dtype=tf.float32),[3,4])
output1 = tf.matmul(input , weight1)
loss1 = tf.reduce_sum(output1 - label1)
weight2 = tf.reshape(tf.range(20,dtype=tf.float32),[4,5])
output2 = tf.matmul(output1 , weight2)
loss2 = tf.reduce_sum(output2 - label2)
grad1 = tf.gradients(loss1,weight1)
grad2 = tf.gradients(loss2,weight2)
grad3 = tf.gradients(loss1+loss2,weight2)
with tf.Session() as sess:
print(sess.run(grad1))
print(sess.run(grad2))
print(sess.run(grad3))
# [array([[1., 1., 1., 1.],
# [2., 2., 2., 2.],
# [3., 3., 3., 3.]], dtype=float32)]
# [array([[32., 32., 32., 32., 32.],
# [38., 38., 38., 38., 38.],
# [44., 44., 44., 44., 44.],
# [50., 50., 50., 50., 50.]], dtype=float32)]
# [array([[32., 32., 32., 32., 32.],
# [38., 38., 38., 38., 38.],
# [44., 44., 44., 44., 44.],
# [50., 50., 50., 50., 50.]], dtype=float32)]
我想我很清楚,我想要一个共同的损失,而不是单独的损失,我已经在我的问题中提出了一个解决方案,我也可以考虑分开的损失和尽量减少它们,这也将是一个坏的解决方案,在我的设置。顺便说一下,当最小化损失1
@user1935724时,您不需要指定变量列表,您应该首先描述共同损失
和损失1
,损失2
之间的数学关系。加法还是乘法?我认为共同损失意味着加法。非常感谢更新!这是否意味着我们真的没有办法先把损失合并起来,然后控制梯度的流动。@user1935724请看我补充的答案
input = tf.constant([[1,2,3]],tf.float32)
label1 = tf.constant([[1,2,3,4]],tf.float32)
label2 = tf.constant([[1,2,3,4,5]],tf.float32)
weight1 = tf.reshape(tf.range(12,dtype=tf.float32),[3,4])
output1 = tf.matmul(input , weight1)
loss1 = tf.reduce_sum(output1 - label1)
weight2 = tf.reshape(tf.range(20,dtype=tf.float32),[4,5])
output2 = tf.matmul(output1 , weight2)
loss2 = tf.reduce_sum(output2 - label2)
grad1 = tf.gradients(loss1,weight1)
grad2 = tf.gradients(loss2,weight2)
grad3 = tf.gradients(loss1+loss2,weight2)
with tf.Session() as sess:
print(sess.run(grad1))
print(sess.run(grad2))
print(sess.run(grad3))
# [array([[1., 1., 1., 1.],
# [2., 2., 2., 2.],
# [3., 3., 3., 3.]], dtype=float32)]
# [array([[32., 32., 32., 32., 32.],
# [38., 38., 38., 38., 38.],
# [44., 44., 44., 44., 44.],
# [50., 50., 50., 50., 50.]], dtype=float32)]
# [array([[32., 32., 32., 32., 32.],
# [38., 38., 38., 38., 38.],
# [44., 44., 44., 44., 44.],
# [50., 50., 50., 50., 50.]], dtype=float32)]