Confusion matrix visualization in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of confusion matrix plot code

What will be the output of the following code snippet that plots a confusion matrix using TensorFlow and Matplotlib?

TensorFlow

import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np

true_labels = [0, 1, 2, 2, 0, 1]
pred_labels = [0, 2, 2, 2, 0, 0]

cm = tf.math.confusion_matrix(true_labels, pred_labels, num_classes=3).numpy()

plt.imshow(cm, interpolation='nearest', cmap=plt.cm.Blues)
plt.title('Confusion Matrix')
plt.colorbar()
plt.xlabel('Predicted label')
plt.ylabel('True label')
plt.xticks(np.arange(3), ['Class 0', 'Class 1', 'Class 2'])
plt.yticks(np.arange(3), ['Class 0', 'Class 1', 'Class 2'])
plt.show()

A A 3x3 heatmap plot showing counts with highest values on diagonal except for class 1

B A 3x3 heatmap plot with all zeros

C A bar chart showing counts of predicted labels

D A line plot showing true vs predicted labels

💡 Hint

Think about how confusion matrices represent counts of true vs predicted labels.

The code computes a confusion matrix for 3 classes and plots it as a heatmap. The matrix shows counts of predictions vs true labels. Since class 1 has some misclassifications, the diagonal is not perfect.

❓ Model Choice

intermediate

Best model output for confusion matrix visualization

You have a multi-class classification problem with 4 classes. Which TensorFlow function will correctly compute the confusion matrix to visualize model performance?

A tf.math.confusion_matrix(labels, predictions, num_classes=4)

B tf.keras.metrics.ConfusionMatrix(labels, predictions)

C tf.confusion_matrix(labels, predictions)

D tf.metrics.confusion_matrix(labels, predictions)

💡 Hint

Check the TensorFlow API for confusion matrix functions.

The correct function is tf.math.confusion_matrix which takes labels and predictions and optional num_classes argument.

❓ Metrics

advanced

Interpreting confusion matrix metrics

Given this confusion matrix for a 3-class problem: [[5, 2, 0], [1, 7, 1], [0, 2, 6]] What is the precision for class 1 (index 1)?

A 7 / (1 + 7 + 1) = 0.78

B 7 / (5 + 1 + 0) = 1.17

C 7 / (2 + 7 + 2) = 0.64

D 7 / (7 + 1 + 2) = 0.70

💡 Hint

Precision = True Positives / (True Positives + False Positives). Look at the predicted column for class 1.

Precision for class 1 is TP / (TP + FP). TP is cm[1][1] = 7. FP is sum of column 1 excluding TP: 2 + 2 = 4. So precision = 7 / (7 + 4) = 7/11 ≈ 0.636, closest to option C's calculation.

🔧 Debug

advanced

Debugging confusion matrix visualization code

What error will this code raise? import tensorflow as tf true = [0, 1, 2] pred = [0, 1] cm = tf.math.confusion_matrix(true, pred) print(cm.numpy())

A TypeError: unsupported operand type(s)

B ValueError: Shapes of true and pred do not match

C IndexError: list index out of range

D No error, prints confusion matrix

💡 Hint

Check if true and pred lists have the same length.

The true and pred lists have different lengths, so TensorFlow raises a ValueError about shape mismatch.

🧠 Conceptual

expert

Choosing visualization method for confusion matrix

You want to visualize a confusion matrix for a 10-class classification problem with highly imbalanced classes. Which approach is best to clearly show the model's performance?

A Plot raw counts heatmap with a color scale from 0 to max count

B Plot a line chart of accuracy over epochs

C Plot a bar chart of total predictions per class

D Plot normalized confusion matrix showing percentages per true class

💡 Hint

Normalization helps compare classes with different sample sizes.

Normalized confusion matrices show proportions per true class, making it easier to interpret performance when classes are imbalanced.