SimpleRNN layer in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output shape of SimpleRNN layer

What is the output shape of the following SimpleRNN layer when given input shape (batch_size=32, timesteps=10, features=8)?

TensorFlow

import tensorflow as tf

rnn = tf.keras.layers.SimpleRNN(16)
input_shape = (32, 10, 8)
inputs = tf.random.uniform(input_shape)
outputs = rnn(inputs)
print(outputs.shape)

A (32, 10, 16)

B (32, 16)

C (10, 16)

D (32, 8)

💡 Hint

SimpleRNN by default returns only the last output for each sequence.

The SimpleRNN layer with units=16 returns the last output for each sequence, so the output shape is (batch_size, units) = (32, 16).

❓ Model Choice

intermediate

Choosing SimpleRNN for sequence data

Which of the following tasks is best suited for a SimpleRNN layer?

A Predicting the next word in a sentence based on previous words

B Classifying images of cats and dogs

C Clustering customer data based on purchase history

D Reducing dimensionality of tabular data

💡 Hint

SimpleRNN is designed to process sequences over time.

SimpleRNN is a recurrent layer that processes sequences, making it suitable for tasks like predicting the next word in a sentence.

❓ Hyperparameter

advanced

Effect of return_sequences=True in SimpleRNN

What is the effect of setting return_sequences=True in a SimpleRNN layer?

A The layer returns the output at each timestep, producing a 3D tensor

B The layer returns the sum of all outputs across timesteps

C The layer returns the input sequence unchanged

D The layer returns only the last output, producing a 2D tensor

💡 Hint

Think about whether the output includes all timesteps or just the last one.

Setting return_sequences=True makes SimpleRNN return the output at every timestep, resulting in a 3D tensor (batch_size, timesteps, units).

❓ Metrics

advanced

Interpreting training loss for SimpleRNN model

A SimpleRNN model for sequence classification shows training loss decreasing steadily but validation loss increasing after some epochs. What does this indicate?

A The model is underfitting both training and validation data

B The model has a bug causing loss to increase

C The model is perfectly generalizing to new data

D The model is overfitting the training data

💡 Hint

Think about what it means when training loss improves but validation loss worsens.

When training loss decreases but validation loss increases, the model is learning training data too well and not generalizing, which is overfitting.

🔧 Debug

expert

Debugging SimpleRNN input shape error

Given this code snippet, what error will occur and why? import tensorflow as tf rnn = tf.keras.layers.SimpleRNN(10) inputs = tf.random.uniform((32, 10)) outputs = rnn(inputs)

TensorFlow

import tensorflow as tf
rnn = tf.keras.layers.SimpleRNN(10)
inputs = tf.random.uniform((32, 10))
outputs = rnn(inputs)

A No error, outputs shape will be (32, 10)

B TypeError because inputs are not integers

C ValueError because input must be 3D but got 2D tensor

D RuntimeError due to missing activation function

💡 Hint

Check the expected input shape for SimpleRNN layers.

SimpleRNN expects input shape (batch_size, timesteps, features). The input here is 2D (batch_size, timesteps) missing features dimension, causing ValueError.