CODEWITHSUNDEEP
pythonmachine-learningneural-networkclassification
jeffs
jeffs

Reputation: 321

Classification using neuronal networks

I have built up a NN for classification, but when trying to compile I get problems with the dimensions of input and output:

from keras.models import Sequential
from keras.layers import Dense

# data splited into input (X) and output (y) variables
model = Sequential()
model.add(Dense(12, input_dim=456, activation='relu'))
model.add(Dense(8, activation='relu'))
model.add(Dense(8, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

Here are the dimensions of my y and X

print(y.shape, X.shape)
(8000, 1) (8000, 456, 3)

I have 8000 sub sets which contain 456 particles(x,y,z); and I have labels which are in y ranging from 0 to 7; this is also why my output layer has 8 nodes.

But when I fit with

model.fit(X, y, epochs=15, batch_size=10)

I do not get why this error occurs:

ValueError: Error when checking input: expected dense_26_input to have 2 dimensions, but got array with shape (8000, 456, 3)

Any suggestions?

Upvotes: 2

Views: 36

Answers (1)

Orphee Faucoz
Orphee Faucoz

Reputation: 1290

To answer your question, you can achieve what you want by doing :

from tensorflow.keras import Sequential
from tensorflow.keras.layers import Dense

model = Sequential()
model.add(Dense(12, input_shape=(456,3), activation='relu'))

model.add(Dense(8, activation='relu'))
model.add(Dense(8, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

model.summary()

Edit:

I think what you're looking for is that type of architecture :

from tensorflow.keras import Sequential
from tensorflow.keras.layers import Dense, Flatten

model = Sequential()
model.add(Dense(12, input_shape=(456,3), activation='relu'))
model.add(Flatten())
model.add(Dense(8, activation='relu'))
model.add(Dense(8, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

model.summary()

So that it only output the 8 labels

Upvotes: 1

Related Questions