Pooling layers (MaxPool, AvgPool) in TensorFlow - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to add a MaxPooling2D layer with pool size 2x2.

TensorFlow

model.add(tf.keras.layers.MaxPooling2D(pool_size=[1]))

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

A(4, 4)

B(2, 2)

C(3, 3)

D(1, 1)

Attempts:

3 left

💡 Hint

Common Mistakes

Using pool_size (1, 1) which does not reduce size.

Confusing pool_size with strides.

The pool_size parameter defines the size of the window to take the maximum over. (2, 2) is a common choice for downsampling.

2fill in blank

medium

Complete the code to add an AveragePooling2D layer with pool size 3x3.

TensorFlow

model.add(tf.keras.layers.AveragePooling2D(pool_size=[1]))

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

A(4, 4)

B(2, 2)

C(1, 1)

D(3, 3)

Attempts:

3 left

💡 Hint

Common Mistakes

Using MaxPooling2D instead of AveragePooling2D.

Using wrong pool_size like (2, 2) instead of (3, 3).

AveragePooling2D with pool_size (3, 3) computes the average over 3x3 windows, reducing spatial dimensions accordingly.

3fill in blank

hard

Fix the error in the code to correctly add a MaxPooling2D layer with stride 2.

TensorFlow

model.add(tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=[1]))

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

A(2, 2)

B(1, 1)

C(3, 3)

D(4, 4)

Attempts:

3 left

💡 Hint

Common Mistakes

Setting strides to (1, 1) which does not downsample.

Using strides larger than pool_size causing skipped regions.

The strides parameter controls how far the pooling window moves each step. For downsampling by half, strides should be (2, 2).

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps each word to its length if length is greater than 3.

TensorFlow

{word: [1] for word in words if [2]

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

Alen(word)

Blen(word) > 3

Cword.startswith('a')

Dword.upper()

Attempts:

3 left

💡 Hint

Common Mistakes

Using word.upper() as value which changes the word instead of length.

Using word.startswith('a') as condition which filters differently.

The dictionary comprehension maps each word to its length (len(word)) only if the length is greater than 3 (len(word) > 3).

5fill in blank

hard

Fill all three blanks to create a dictionary comprehension mapping uppercase words to their lengths if length is greater than 4.

TensorFlow

{ [1]: [2] for word in words if [3] }

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

Aword.upper()

Blen(word)

Clen(word) > 4

Dword.lower()

Attempts:

3 left

💡 Hint

Common Mistakes

Using word.lower() as key instead of uppercase.

Using len(word) < 4 as condition which filters incorrectly.

The dictionary comprehension maps the uppercase version of each word (word.upper()) to its length (len(word)) only if the length is greater than 4 (len(word) > 4).

Practice

1. What is the main purpose of a MaxPooling layer in a neural network?

easy

A. To reduce the size of the input while keeping the strongest features

B. To increase the size of the input by adding more pixels

C. To randomly change pixel values for data augmentation

D. To convert images into grayscale

Solution

1. Step 1: Understand MaxPooling function

   MaxPooling scans small regions of the input and selects the maximum value, reducing size but keeping strong signals.

2. Step 2: Compare options with MaxPooling role

   Only To reduce the size of the input while keeping the strongest features describes reducing size while keeping strongest features, matching MaxPooling's purpose.

3. Final Answer:

   To reduce the size of the input while keeping the strongest features -> Option A

4. Quick Check:

   MaxPooling = reduce size + keep max [OK]

Hint: MaxPooling keeps max values, shrinking input size [OK]

Common Mistakes:

• Confusing MaxPooling with data augmentation
• Thinking MaxPooling increases input size
• Mixing MaxPooling with color conversion

2. Which of the following is the correct way to create a 2x2 MaxPooling layer with stride 2 in TensorFlow?

easy

A. tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=2)

B. tf.keras.layers.MaxPooling2D(pool_size=2, stride=2)

C. tf.keras.layers.MaxPooling2D(pool=(2, 2), strides=(2, 2))

D. tf.keras.layers.MaxPooling2D(size=2, stride=2)

Solution

1. Step 1: Recall correct TensorFlow MaxPooling2D syntax

   The correct parameters are pool_size and strides, both can be tuples or integers.

2. Step 2: Check each option's parameter names and values

   tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=2) uses pool_size=(2, 2) and strides=2, which is valid. Others use wrong parameter names or formats.

3. Final Answer:

   tf.keras.layers.MaxPooling2D(pool_size=(2, 2), strides=2) -> Option A

4. Quick Check:

   Correct params: pool_size, strides [OK]

Hint: Use pool_size and strides parameters exactly [OK]

Common Mistakes:

• Using 'stride' instead of 'strides'
• Using 'pool' or 'size' instead of 'pool_size'
• Passing integers without tuple when tuple needed

3. Given the input tensor below, what is the output shape after applying a 2x2 MaxPooling layer with stride 2?

input_tensor = tf.constant([[[[1], [2], [3], [4]],
                             [[5], [6], [7], [8]],
                             [[9], [10], [11], [12]],
                             [[13], [14], [15], [16]]]], dtype=tf.float32)

medium

A. (1, 3, 3, 1)

B. (1, 4, 4, 1)

C. (1, 2, 2, 1)

D. (1, 1, 1, 1)

Solution

1. Step 1: Understand input shape and pooling parameters

   Input shape is (1, 4, 4, 1). Pool size is 2x2, stride 2, so output spatial dims shrink by factor 2.

2. Step 2: Calculate output shape after pooling

   Height and width: floor(4/2) = 2. Channels and batch size remain same. Output shape is (1, 2, 2, 1).

3. Final Answer:

   (1, 2, 2, 1) -> Option C

4. Quick Check:

   Pooling halves spatial dims = (1, 2, 2, 1) [OK]

Hint: Output dims = input dims divided by stride [OK]

Common Mistakes:

• Forgetting batch and channel dimensions
• Using input size instead of divided size
• Confusing stride with pool size

4. You wrote this code but get an error:

layer = tf.keras.layers.AvgPooling2D(pool_size=2, strides=2)
output = layer(input_tensor)

What is the most likely cause of the error?

medium

A. input_tensor must be a numpy array, not a tensor

B. pool_size must be a tuple, not an integer

C. AvgPooling2D does not accept strides parameter

D. input_tensor is missing batch dimension

Solution

1. Step 1: Check input_tensor shape requirements

   Pooling layers expect 4D input: (batch, height, width, channels). Missing batch dimension causes error.

2. Step 2: Verify other options

   pool_size can be int or tuple, strides is accepted, and input can be tensor. So only missing batch dimension is likely cause.

3. Final Answer:

   input_tensor is missing batch dimension -> Option D

4. Quick Check:

   Pooling needs 4D input [OK]

Hint: Pooling layers require 4D input tensors [OK]

Common Mistakes:

• Passing 3D tensor without batch size
• Thinking strides parameter is invalid
• Confusing tensor and numpy array input

5. You want to reduce noise in feature maps but keep important edges. Which pooling strategy should you use and why? MaxPooling or AveragePooling?

hard

A. Use MaxPooling because it keeps strongest signals like edges

B. Use AveragePooling because it smooths out noise by averaging

C. Use MaxPooling because it averages values to reduce noise

D. Use AveragePooling because it selects max values to keep edges

Solution

1. Step 1: Understand effect of MaxPooling and AveragePooling

   MaxPooling keeps max values, emphasizing strong signals like edges but may keep noise spikes. AveragePooling smooths values, reducing noise.

2. Step 2: Match goal with pooling type

   To reduce noise but keep edges, smoothing by AveragePooling is better. MaxPooling keeps edges but may keep noise.

3. Final Answer:

   Use AveragePooling because it smooths out noise by averaging -> Option B

4. Quick Check:

   Noise reduction = AveragePooling [OK]

Hint: AveragePooling smooths noise; MaxPooling keeps edges [OK]

Common Mistakes:

• Confusing max with average effects
• Thinking MaxPooling reduces noise
• Mixing pooling effects on edges and noise