Sequential model API in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of a simple Sequential model prediction

What is the output shape of the prediction from this Sequential model when given input shape (None, 10)?

TensorFlow

import tensorflow as tf
model = tf.keras.Sequential([
    tf.keras.layers.Dense(5, activation='relu', input_shape=(10,)),
    tf.keras.layers.Dense(3, activation='softmax')
])
import numpy as np
sample_input = np.random.random((1, 10))
prediction = model(sample_input)
prediction_shape = prediction.shape
print(prediction_shape)

A (1, 3)

B (10, 3)

C (1, 5)

D (3,)

💡 Hint

Remember the output shape depends on the last layer's units and batch size.

The model input shape is (None, 10), batch size 1 means input is (1,10). The last Dense layer has 3 units, so output shape is (1,3).

❓ Model Choice

intermediate

Choosing the correct Sequential model for binary classification

Which Sequential model architecture is best suited for a binary classification problem with input features of size 20?

A Sequential([Dense(1, activation='relu', input_shape=(20,)), Dense(1, activation='sigmoid')])

B Sequential([Dense(10, activation='relu', input_shape=(20,)), Dense(3, activation='softmax')])

C Sequential([Dense(10, activation='relu', input_shape=(20,)), Dense(1, activation='sigmoid')])

D Sequential([Dense(20, activation='relu'), Dense(1, activation='softmax')])

💡 Hint

Binary classification needs one output neuron with sigmoid activation.

For binary classification, the output layer should have 1 unit with sigmoid activation to output probabilities between 0 and 1.

❓ Hyperparameter

advanced

Effect of changing batch size in Sequential model training

What is the most likely effect of increasing the batch size from 16 to 256 during training of a Sequential model?

A Training will use less memory and always converge slower.

B Training will use less memory and produce more accurate models.

C Training will use more memory and always overfit the data.

D Training will use more memory and may converge faster but with less noisy gradient updates.

💡 Hint

Think about how batch size affects memory and gradient noise.

Larger batch sizes require more memory but produce smoother gradient estimates, which can speed up convergence but may reduce generalization.

❓ Metrics

advanced

Correct metric for multi-class classification with Sequential model

Which metric should be used to evaluate a Sequential model trained for 4-class classification?

A mean_absolute_error

B accuracy

C mean_squared_error

D binary_crossentropy

💡 Hint

Think about the type of problem and metric suitability.

Accuracy is the standard metric for multi-class classification to measure correct predictions ratio.

🔧 Debug

expert

Identify the error in this Sequential model code

What error will this code raise when run?

TensorFlow

import tensorflow as tf
model = tf.keras.Sequential()
model.add(tf.keras.layers.Dense(10, activation='relu'))
model.add(tf.keras.layers.Dense(1, activation='sigmoid'))
model.compile(optimizer='adam', loss='binary_crossentropy')
import numpy as np
x = np.random.random((5, 20))
y = np.random.randint(0, 2, size=(5, 1))
model.fit(x, y, epochs=1)

A No error, code runs successfully

B TypeError: optimizer argument must be a string

C RuntimeError: model.fit requires validation data

D ValueError: Input shape is not defined for the first layer

💡 Hint

Keras Sequential models can infer input shape from training data.

Keras Sequential models infer the input shape automatically from the input data during the first forward pass in fit(), so no error is raised.