Keras as TensorFlow's high-level API - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Model Choice

intermediate

Choosing the correct Keras model for image classification

You want to build a simple image classifier using Keras in TensorFlow. Which model type is best suited for this task?

A Sequential model with Conv2D and Dense layers

B Functional API model with only Dense layers

C Sequential model with only LSTM layers

D Functional API model with Embedding layers only

💡 Hint

Think about the type of data images are and which layers process spatial information.

Images have spatial structure best captured by convolutional layers (Conv2D). A Sequential model with Conv2D and Dense layers is a simple and effective choice for image classification.

❓ Predict Output

intermediate

Output shape of a Keras Conv2D layer

What is the output shape of this Keras layer when input shape is (32, 32, 3)?

TensorFlow

from tensorflow.keras.layers import Conv2D
layer = Conv2D(filters=16, kernel_size=3, strides=1, padding='valid')
output_shape = layer.compute_output_shape((None, 32, 32, 3))
print(output_shape)

A (None, 32, 32, 16)

B (None, 31, 31, 16)

C (None, 30, 30, 16)

D (None, 30, 30, 3)

💡 Hint

Padding='valid' means no padding, so output size shrinks by kernel_size - 1.

With kernel size 3 and valid padding, output height and width = input - 3 + 1 = 30. Filters define output channels = 16.

❓ Hyperparameter

advanced

Choosing the right optimizer for faster convergence

You train a Keras model but it learns very slowly. Which optimizer change is most likely to speed up training without losing stability?

A Switch from Adam to SGD optimizer

B Switch from SGD to Adam optimizer

C Switch from RMSprop to SGD optimizer with momentum=0

D Switch from Adam to RMSprop with learning rate 10 times smaller

💡 Hint

Adam adapts learning rates per parameter and often converges faster than vanilla SGD.

Adam optimizer combines momentum and adaptive learning rates, usually speeding up convergence compared to SGD.

❓ Metrics

advanced

Interpreting Keras model accuracy during training

After training a Keras classification model, the training accuracy is 95% but validation accuracy is 70%. What does this indicate?

A The model is overfitting the training data

B The model is underfitting the training data

C The model has perfect generalization

D The validation data is corrupted

💡 Hint

High training accuracy but much lower validation accuracy usually means the model memorizes training data.

Overfitting happens when the model learns training data too well but fails to generalize to new data, causing lower validation accuracy.

🔧 Debug

expert

Identifying the cause of a Keras model training error

You run this Keras code and get a ValueError: 'Input 0 of layer dense_1 is incompatible with the layer: expected axis -1 of input shape to have value 64 but received input with shape (None, 32)'. What is the most likely cause?

TensorFlow

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten
model = Sequential([
  Flatten(input_shape=(8,8)),
  Dense(64, activation='relu'),
  Dense(10, activation='softmax')
])
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy')
model.fit(x_train, y_train, epochs=5)

A The input_shape in Flatten is incorrect; it should be (64,) instead of (8,8)

B The Dense layer expects input with last dimension 64, but Flatten outputs (None, 64), so the error is in the Dense layer

C The model is missing an input layer, causing shape mismatch

D The Flatten layer output shape is (None, 64), but Dense expects input with last dimension 64, so input data shape is wrong

💡 Hint

Flatten converts (8,8) to 64 features. The error says input to Dense has shape (None, 32), which is unexpected.

The error means the data fed to the model has shape (None, 8, 4) or similar, not (None, 8, 8). Flatten outputs 64 features, but input data shape is inconsistent.