Time series with RNN in TensorFlow - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to create a simple RNN layer for time series data.

TensorFlow

model = tf.keras.Sequential([
    tf.keras.layers.SimpleRNN([1], input_shape=(10, 1))
])

Drag options to blanks, or click blank then click option'

A32

B0.1

Crelu

D10

Attempts:

3 left

💡 Hint

Common Mistakes

Using an activation function or dropout rate instead of units.

Setting input_shape as the number of units.

The number 32 specifies the number of units (neurons) in the SimpleRNN layer, which is required to define the layer size.

2fill in blank

medium

Complete the code to compile the RNN model with mean squared error loss.

TensorFlow

model.compile(optimizer='adam', loss=[1], metrics=['mae'])

Drag options to blanks, or click blank then click option'

A'accuracy'

B'binary_crossentropy'

C'categorical_crossentropy'

D'mse'

Attempts:

3 left

💡 Hint

Common Mistakes

Using classification loss functions like 'accuracy' or 'categorical_crossentropy'.

For time series regression, mean squared error ('mse') is the appropriate loss function.

3fill in blank

hard

Fix the error in the code to correctly reshape the input data for the RNN.

TensorFlow

X_train = X_train.reshape((X_train.shape[0], [1], 1))

Drag options to blanks, or click blank then click option'

AX_train.shape[1]

BX_train.shape[0]

C1

DX_train.shape[2]

Attempts:

3 left

💡 Hint

Common Mistakes

Using the batch size or feature dimension incorrectly.

Setting time steps to 1 which loses sequence information.

The second dimension should be the number of time steps, which is the original second dimension of X_train.

4fill in blank

hard

Fill both blanks to add a Dense output layer for univariate time series regression.

TensorFlow

tf.keras.layers.Dense([1], activation=[2])

Drag options to blanks, or click blank then click option'

A1

B'linear'

C32

D'relu'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'relu' activation which clips negative values.

Setting units to hidden size like 32 instead of output size.

Dense layer with 1 unit and linear activation is standard for predicting a single continuous value in time series regression.

5fill in blank

hard

Fill all three blanks to train the RNN model with standard hyperparameters.

TensorFlow

history = model.fit(X_train, y_train, epochs=[1], batch_size=[2], validation_split=[3])

Drag options to blanks, or click blank then click option'

A50

B32

C0.2

D10

Attempts:

3 left

💡 Hint

Common Mistakes

Using too few epochs like 10.

Omitting validation_split or using invalid values.

Using 50 epochs, batch_size=32, and validation_split=0.2 are common settings for training RNN models on time series data.