问题
So I'm trying to create an autoencoder that will take text reviews and find a lower dimensional representation. I'm using keras and I want my loss function to compare the output of the AE to the output of the embedding layer. Unfortunately, it gives me the following error. I'm pretty sure the problem is with my loss function but I can't seem to resolve the issue.
Autoencoder
print X_train.shape
input_i = Input(shape=(200,))
embedding = Embedding(input_dim=weights.shape[0],output_dim=weights.shape[1],
weights=[weights])(input_i)
encoded_h1 = Dense(64, activation='tanh')(embedding)
encoded_h2 = Dense(32, activation='tanh')(encoded_h1)
encoded_h3 = Dense(16, activation='tanh')(encoded_h2)
encoded_h4 = Dense(8, activation='tanh')(encoded_h3)
encoded_h5 = Dense(4, activation='tanh')(encoded_h4)
latent = Dense(2, activation='tanh')(encoded_h5)
decoder_h1 = Dense(4, activation='tanh')(latent)
decoder_h2 = Dense(8, activation='tanh')(decoder_h1)
decoder_h3 = Dense(16, activation='tanh')(decoder_h2)
decoder_h4 = Dense(32, activation='tanh')(decoder_h3)
decoder_h5 = Dense(64, activation='tanh')(decoder_h4)
output = Dense(weights.shape[1], activation='tanh')(decoder_h5)
autoencoder = Model(input_i,output)
encoder = Model(input_i,latent)
print autoencoder.summary()
import keras.backend as K
import tensorflow as tf
def embedded_mse(x_true, e_pred):
print output
print embedding
mse = K.mean(K.square(output - embedding))
print mse
return tf.Session().run(mse)
autoencoder.compile(optimizer='adadelta',
loss=embedded_mse)
autoencoder.fit(X_train,X_train,epochs=10,
batch_size=256, validation_split=.1)
Output
(100000, 200)
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_47 (InputLayer) (None, 200) 0
_________________________________________________________________
embedding_31 (Embedding) (None, 200, 100) 21833700
_________________________________________________________________
dense_528 (Dense) (None, 200, 64) 6464
_________________________________________________________________
dense_529 (Dense) (None, 200, 32) 2080
_________________________________________________________________
dense_530 (Dense) (None, 200, 16) 528
_________________________________________________________________
dense_531 (Dense) (None, 200, 8) 136
_________________________________________________________________
dense_532 (Dense) (None, 200, 4) 36
_________________________________________________________________
dense_533 (Dense) (None, 200, 2) 10
_________________________________________________________________
dense_534 (Dense) (None, 200, 4) 12
_________________________________________________________________
dense_535 (Dense) (None, 200, 8) 40
_________________________________________________________________
dense_536 (Dense) (None, 200, 16) 144
_________________________________________________________________
dense_537 (Dense) (None, 200, 32) 544
_________________________________________________________________
dense_538 (Dense) (None, 200, 64) 2112
_________________________________________________________________
dense_539 (Dense) (None, 200, 100) 6500
=================================================================
Total params: 21,852,306
Trainable params: 21,852,306
Non-trainable params: 0
_________________________________________________________________
None
Tensor("dense_539/Tanh:0", shape=(?, 200, 100), dtype=float32)
Tensor("embedding_31/Gather:0", shape=(?, 200, 100), dtype=float32)
Tensor("loss_48/dense_539_loss/Mean:0", shape=(), dtype=float32)
Error
---------------------------------------------------------------------------
InvalidArgumentError Traceback (most recent call last)
<ipython-input-155-a18e0c32f59b> in <module>()
1 autoencoder.compile(optimizer='adadelta',
----> 2 loss=embedded_mse)
3 autoencoder.fit(X_train,embedding,epochs=10,
4 batch_size=256, validation_split=.1)
/home/andrew/.local/lib/python2.7/site-packages/keras/engine/training.pyc in compile(self, optimizer, loss, metrics, loss_weights, sample_weight_mode, weighted_metrics, target_tensors, **kwargs)
848 with K.name_scope(self.output_names[i] + '_loss'):
849 output_loss = weighted_loss(y_true, y_pred,
--> 850 sample_weight, mask)
851 if len(self.outputs) > 1:
852 self.metrics_tensors.append(output_loss)
/home/andrew/.local/lib/python2.7/site-packages/keras/engine/training.pyc in weighted(y_true, y_pred, weights, mask)
448 """
449 # score_array has ndim >= 2
--> 450 score_array = fn(y_true, y_pred)
451 if mask is not None:
452 # Cast the mask to floatX to avoid float64 upcasting in theano
<ipython-input-153-73211fc383a5> in embedded_mse(x_true, e_pred)
7 print mse
8
----> 9 return tf.Session().run(mse)
/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in run(self, fetches, feed_dict, options, run_metadata)
893 try:
894 result = self._run(None, fetches, feed_dict, options_ptr,
--> 895 run_metadata_ptr)
896 if run_metadata:
897 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in _run(self, handle, fetches, feed_dict, options, run_metadata)
1122 if final_fetches or final_targets or (handle and feed_dict_tensor):
1123 results = self._do_run(handle, final_targets, final_fetches,
-> 1124 feed_dict_tensor, options, run_metadata)
1125 else:
1126 results = []
/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in _do_run(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)
1319 if handle is None:
1320 return self._do_call(_run_fn, self._session, feeds, fetches, targets,
-> 1321 options, run_metadata)
1322 else:
1323 return self._do_call(_prun_fn, self._session, handle, feeds, fetches)
/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/client/session.pyc in _do_call(self, fn, *args)
1338 except KeyError:
1339 pass
-> 1340 raise type(e)(node_def, op, message)
1341
1342 def _extend_graph(self):
InvalidArgumentError: You must feed a value for placeholder tensor 'input_47' with dtype float and shape [?,200]
[[Node: input_47 = Placeholder[dtype=DT_FLOAT, shape=[?,200], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
Caused by op u'input_47', defined at:
File "/usr/lib/python2.7/runpy.py", line 174, in _run_module_as_main
"__main__", fname, loader, pkg_name)
File "/usr/lib/python2.7/runpy.py", line 72, in _run_code
exec code in run_globals
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel_launcher.py", line 16, in <module>
app.launch_new_instance()
File "/home/andrew/.local/lib/python2.7/site-packages/traitlets/config/application.py", line 658, in launch_instance
app.start()
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel/kernelapp.py", line 477, in start
ioloop.IOLoop.instance().start()
File "/home/andrew/.local/lib/python2.7/site-packages/zmq/eventloop/ioloop.py", line 177, in start
super(ZMQIOLoop, self).start()
File "/home/andrew/.local/lib/python2.7/site-packages/tornado/ioloop.py", line 888, in start
handler_func(fd_obj, events)
File "/home/andrew/.local/lib/python2.7/site-packages/tornado/stack_context.py", line 277, in null_wrapper
return fn(*args, **kwargs)
File "/home/andrew/.local/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py", line 440, in _handle_events
self._handle_recv()
File "/home/andrew/.local/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py", line 472, in _handle_recv
self._run_callback(callback, msg)
File "/home/andrew/.local/lib/python2.7/site-packages/zmq/eventloop/zmqstream.py", line 414, in _run_callback
callback(*args, **kwargs)
File "/home/andrew/.local/lib/python2.7/site-packages/tornado/stack_context.py", line 277, in null_wrapper
return fn(*args, **kwargs)
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel/kernelbase.py", line 283, in dispatcher
return self.dispatch_shell(stream, msg)
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel/kernelbase.py", line 235, in dispatch_shell
handler(stream, idents, msg)
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel/kernelbase.py", line 399, in execute_request
user_expressions, allow_stdin)
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel/ipkernel.py", line 196, in do_execute
res = shell.run_cell(code, store_history=store_history, silent=silent)
File "/home/andrew/.local/lib/python2.7/site-packages/ipykernel/zmqshell.py", line 533, in run_cell
return super(ZMQInteractiveShell, self).run_cell(*args, **kwargs)
File "/home/andrew/.local/lib/python2.7/site-packages/IPython/core/interactiveshell.py", line 2718, in run_cell
interactivity=interactivity, compiler=compiler, result=result)
File "/home/andrew/.local/lib/python2.7/site-packages/IPython/core/interactiveshell.py", line 2822, in run_ast_nodes
if self.run_code(code, result):
File "/home/andrew/.local/lib/python2.7/site-packages/IPython/core/interactiveshell.py", line 2882, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "<ipython-input-152-7732fda181fc>", line 2, in <module>
input_i = Input(shape=(200,))
File "/home/andrew/.local/lib/python2.7/site-packages/keras/engine/topology.py", line 1436, in Input
input_tensor=tensor)
File "/home/andrew/.local/lib/python2.7/site-packages/keras/legacy/interfaces.py", line 87, in wrapper
return func(*args, **kwargs)
File "/home/andrew/.local/lib/python2.7/site-packages/keras/engine/topology.py", line 1347, in __init__
name=self.name)
File "/home/andrew/.local/lib/python2.7/site-packages/keras/backend/tensorflow_backend.py", line 442, in placeholder
x = tf.placeholder(dtype, shape=shape, name=name)
File "/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py", line 1548, in placeholder
return gen_array_ops._placeholder(dtype=dtype, shape=shape, name=name)
File "/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 2094, in _placeholder
name=name)
File "/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
op_def=op_def)
File "/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2630, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/home/andrew/.local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1204, in __init__
self._traceback = self._graph._extract_stack() # pylint: disable=protected-access
InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'input_47' with dtype float and shape [?,200]
[[Node: input_47 = Placeholder[dtype=DT_FLOAT, shape=[?,200], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
回答1:
There are some issues with your question (e.g. what is weights, used in the Embedding & final Dense layers arguments?). Still, I think a simpler approach would be to disentangle the embedding and the autoencoding parts (they are independent), by building first a simple embedding model and then use its outputs (with predict) to feed your autoencoder. This way you don't have to define a custom loss (BTW, print statements in such functions are not a good idea).
Without knowing the details of your data, the following 2 models compile OK:
Embedding model (quick adaptation from the docs)
model = Sequential()
model.add(Embedding(1000, 64))
model.compile('rmsprop', 'mse')
Autoencoder:
input_i = Input(shape=(200,100))
encoded_h1 = Dense(64, activation='tanh')(input_i)
encoded_h2 = Dense(32, activation='tanh')(encoded_h1)
encoded_h3 = Dense(16, activation='tanh')(encoded_h2)
encoded_h4 = Dense(8, activation='tanh')(encoded_h3)
encoded_h5 = Dense(4, activation='tanh')(encoded_h4)
latent = Dense(2, activation='tanh')(encoded_h5)
decoder_h1 = Dense(4, activation='tanh')(latent)
decoder_h2 = Dense(8, activation='tanh')(decoder_h1)
decoder_h3 = Dense(16, activation='tanh')(decoder_h2)
decoder_h4 = Dense(32, activation='tanh')(decoder_h3)
decoder_h5 = Dense(64, activation='tanh')(decoder_h4)
output = Dense(100, activation='tanh')(decoder_h5)
autoencoder = Model(input_i,output)
autoencoder.compile('adadelta','mse')
After adapting the above models parameters to your case, this should work fine:
X_embedded = model.predict(X_train)
autoencoder.fit(X_embedded,X_embedded,epochs=10,
batch_size=256, validation_split=.1)
来源：https://stackoverflow.com/questions/47735205/keras-autoencoder-for-text-analysis