Confusion matrix analysis in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Confusion matrix shape from TensorFlow predictions

Given a binary classification model in TensorFlow, what is the shape of the confusion matrix returned by tf.math.confusion_matrix when comparing true labels and predicted labels?

TensorFlow

import tensorflow as tf
true_labels = tf.constant([0, 1, 0, 1, 1])
pred_labels = tf.constant([0, 0, 0, 1, 1])
cm = tf.math.confusion_matrix(true_labels, pred_labels)
cm_shape = cm.shape
print(cm_shape)

A (1, 2)

B (5, 5)

C (3, 3)

D (2, 2)

💡 Hint

The confusion matrix size depends on the number of classes, not the number of samples.

For binary classification, the confusion matrix is 2x2 because there are two classes: 0 and 1.

❓ Metrics

intermediate

Calculating accuracy from confusion matrix

Given the confusion matrix below, what is the accuracy of the model? [[50, 10], [5, 35]]

A 0.80

B 0.85

C 0.90

D 0.75

💡 Hint

Accuracy = (True Positives + True Negatives) / Total samples

Accuracy = (50 + 35) / (50 + 10 + 5 + 35) = 85 / 100 = 0.85

❓ Model Choice

advanced

Choosing model type based on confusion matrix imbalance

You have a confusion matrix with many false negatives but few false positives. Which model adjustment is best to reduce false negatives?

A Increase recall by adjusting the classification threshold lower

B Increase precision by adjusting the classification threshold higher

C Use a simpler model to reduce overfitting

D Use early stopping during training

💡 Hint

Recall focuses on reducing false negatives.

Lowering the classification threshold increases recall, reducing false negatives.

🔧 Debug

advanced

Identifying error in confusion matrix calculation code

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

TensorFlow

import tensorflow as tf
true = tf.constant([0, 1, 2])
pred = tf.constant([0, 1])
cm = tf.math.confusion_matrix(true, pred)
print(cm)

A ValueError due to invalid label values

B TypeError due to wrong input types

C InvalidArgumentError due to shape mismatch

D No error, prints confusion matrix

💡 Hint

True and predicted labels must have the same length.

The true and predicted label tensors have different lengths, causing an InvalidArgumentError.

🧠 Conceptual

expert

Interpreting confusion matrix for multi-class classification

In a 3-class classification problem, the confusion matrix is: [[30, 2, 3], [4, 25, 1], [5, 0, 35]] Which class has the highest precision?

A Class 1

B Class 2

C Class 0

D All classes have equal precision

💡 Hint

Precision = True Positives / (True Positives + False Positives) for each class.

Class 1 precision = 25 / (2 + 25 + 0) = 25 / 27 ≈ 0.926, which is higher than others.