Training lstm model?

Question

trying to train LSTM model with (5,xxx) data stream fe

(5, 17516)
array([[ 3820.,  2873.,  2369., ..., 18865., 16893., 14242.],
   [ 4656.,  3820.,  2873., ..., 19967., 18865., 16893.],
   [ 6210.,  4656.,  3820., ..., 20223., 19967., 18865.],
   [ 8127.,  6210.,  4656., ..., 20319., 20223., 19967.],
   [10844.,  8127.,  6210., ..., 17246., 20319., 20223.]])

here is the model :

def lstm_model(self, window=5):
    self.model = Sequential()
    self.model.add(LSTM(4, input_shape=( window, 1)))
    self.model.add(Dense(1))
    self.model.compile(loss='mean_squared_error', optimizer='adam')
    return self.model

here is the fit :

self.history = self.model.fit(
        windowed_data , self.data.data,
        validation_split=0.2, nb_epoch=55, batch_size=10, verbose=1)

here is the error I'm getting :

ValueError: Error when checking input: expected lstm_6_input to have 3 dimensions, but got array with shape (5, 17516)

What I'm doing wrong ?

---

This seems to solve it.

w.reshape(w.shape[0], w.shape[1],1)

Answer 1

according to keras docs the input data should be a 3d tensor ie (nb_samples, timesteps, input_dim). this is a good tutorial on how to reshape your data for lstm models.