Keras.layers.concatenate generates an error'

Question

I am trying to train a CNN with two input branches. And these two branches (b1, b2) are to be merged into a densely connected layer of 256 neurons with dropout rate of 0.25. This is what I have so far:

batch_size, epochs = 32, 3
ksize = 2
l2_lambda = 0.0001


### My first model(b1)
b1 = Sequential()
b1.add(Conv1D(128*2, kernel_size=ksize,
             activation='relu',
             input_shape=( xtest.shape[1], xtest.shape[2]),
             kernel_regularizer=keras.regularizers.l2(l2_lambda)))
b1.add(Conv1D(128*2, kernel_size=ksize, activation='relu',kernel_regularizer=keras.regularizers.l2(l2_lambda)))
b1.add(MaxPooling1D(pool_size=ksize))
b1.add(Dropout(0.2))

b1.add(Conv1D(128*2, kernel_size=ksize, activation='relu',kernel_regularizer=keras.regularizers.l2(l2_lambda)))
b1.add(MaxPooling1D(pool_size=ksize))
b1.add(Dropout(0.2))

b1.add(Flatten())

###My second model (b2)

b2 = Sequential()
b2.add(Dense(64, input_shape = (5000,), activation='relu',kernel_regularizer=keras.regularizers.l2(l2_lambda)))
b2.add(Dropout(0.1))


##Merging the two models
model = Sequential()
model.add(concatenate([b1, b2],axis = -1))
model.add(Dense(256, activation='relu', kernel_initializer='normal',kernel_regularizer=keras.regularizers.l2(l2_lambda)))
model.add(Dropout(0.25))
model.add(Dense(num_classes, activation='softmax'))

But when I concatenate it gives me the following error:

I first tried using the following command:

  model.add(Merge([b1, b2], mode = 'concat'))

But I got the error that 'ImportError: cannot import name 'Merge''. I am using keras 2.2.2 and python 3.6.

Yes, I tried that too but I got the same error as before. And I am not sure how to change it to functional API. I am new to Keras and machine learning. — Shafa Haider
– Shafa Haider, Commented Sep 14, 2018 at 13:49
Aha! Use b1.output and b2.output instead of b1 and b2. — today
– today, Commented Sep 14, 2018 at 14:11
I just tried it and I got the following error now: TypeError: The added layer must be an instance of class Layer. Found: Tensor("concatenate_1/concat:0", shape=(?, ?), dtype=float32) — Shafa Haider
– Shafa Haider, Commented Sep 14, 2018 at 14:39
I'm not sure if this will help. But xtest.shape[1] is 5000 and xtest.shape[2] is 208. — Shafa Haider
– Shafa Haider, Commented Sep 14, 2018 at 14:41
Sorry! My initial comment was a bit wrong (as a result I deleted it to prevent further confusion). Please take a look at my answer. — today
– today, Commented Sep 14, 2018 at 15:07

today · Accepted Answer · 2018-09-14 15:14:12Z

1

You need to use the functional API to achieve what you are looking for. You can either use Concatenate layer or its equivalent functional API concatenate:

concat = Concatenate(axis=-1)([b1.output, b2.output])
# or you can use the functional api as follows:
#concat = concatenate([b1.output, b2.output], axis=-1)

x = Dense(256, activation='relu', kernel_initializer='normal',
          kernel_regularizer=keras.regularizers.l2(l2_lambda))(concat)
x = Dropout(0.25)(x)
output = Dense(num_classes, activation='softmax')(x)

model = Model([b1.input, b2.input], [output])

Note that I have only converted the last part of your model to functional form. You can do the same thing for the other two models b1 and b2 (actually, it seems that the architecture you are trying to define is one single model that consists of two branches that are merged together). At the end, use model.summary() to see and recheck the architecture of the model.

edited Sep 14, 2018 at 15:14

answered Sep 14, 2018 at 15:06

today

33.6k8 gold badges100 silver badges122 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Shafa Haider Over a year ago

Thank you. My network works fine now. You saved the day for me :)

Collectives™ on Stack Overflow

Keras.layers.concatenate generates an error'

1 Answer 1

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related