CODEWITHSUNDEEP
pythonnumpytensorflowneural-networkkeras
Ggs
Ggs

Reputation: 81

Keras input shape error

I'm new to Keras and I am trying to create my network which needs to learn on a card game. It takes 93 binary inputs with a hidden layer with 40 neurons and a single output neuron which computes a score (from 0 to 25).

model = Sequential()
model.add(Dense(input_dim=93, units=40, activation="sigmoid"))
model.add(Dense(units=2, activation="linear"))
sgd = optimizers.SGD(lr=0.01, clipvalue=0.5)
model.compile(loss="mse", optimizer=sgd, learning_rate=0.01)

I'm trying first to compute (do a forward propagation) of the 93 inputs

this is "s.toInputs()"

[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 1 0 1 1 0 1 1 1 1 0 1 1 1 1 0 1 1 1 1 0 1 0 0 1 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 1 1 0 0 0 0 1 0 1 0 0 0 0 0 1 1 0 0 1 1 0 0 1 0 0 0 0 1 1 0 1]

model.predict(np.array(s.toInputs())

but i get the error

ValueError: Error when checking : expected dense_1_input to have shape (None, 93) but got array with shape (93, 1)

How do I pass the correct parameters?

Upvotes: 5

Views: 5710

Answers (3)

shiva prasad reddy puchala
shiva prasad reddy puchala

Reputation: 29

The reshaping thing works well. Reshape the array to (batch_size, your_input_dimensions)

np.reshape(batch_size, input_dimesions)

Upvotes: -1

Wilmar van Ommeren
Wilmar van Ommeren

Reputation: 7699

Actually s.toInputs() should look like this

[[0,0,0, etc...],[0,1,0, etc...]]

Basically you have to have an array with the following shape: (n_batches, n_attributes)

You have 93 attributes so this should do the trick if you are using tensorflow

np.array(s.toInputs()).reshape(-1, 93)

Fully working example

model = Sequential()
model.add(Dense(input_dim=93, units=40, activation="sigmoid"))
model.add(Dense(units=2, activation="linear"))
sgd = optimizers.SGD(lr=0.01, clipvalue=0.5)
model.compile(loss="mse", optimizer=sgd, learning_rate=0.01)

# random data
n_batches = 10
data = np.random.randint(0,2,93*n_batches)
data = data.reshape(-1,93)

model.predict(data)

Upvotes: 5

sietschie
sietschie

Reputation: 7553

The error message tells you, that your data needs to have the shape (None, 93) (Nonehere means, this dimension can have an arbitrary value. It is the number of your samples)

But your input data has the shape (93,1). Note that the dimensions are reversed. You can use transpose to get your data in the right shape:

model.predict(np.array(s.toInputs()).T)

Upvotes: 1

Related Questions