Generative vs discriminative models in Prompt Engineering / GenAI - Practice Questions

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding model types: Generative vs Discriminative

Which statement correctly describes the main difference between generative and discriminative models?

A Generative models only classify data, whereas discriminative models generate new data samples.

B Discriminative models learn the joint probability of data and labels, while generative models learn the conditional probability of labels given data.

C Generative models learn the joint probability of data and labels, while discriminative models learn the conditional probability of labels given data.

D Generative models always perform better than discriminative models on classification tasks.

💡 Hint

Think about what each model tries to understand about the data and labels.

Generative models learn how data and labels come together (joint probability), enabling them to generate new data. Discriminative models focus on predicting labels from data (conditional probability).

❓ Model Choice

intermediate

Choosing the right model for generating images

You want to create a model that can generate realistic images of cats. Which type of model should you choose?

A Discriminative model like decision trees

B Discriminative model like logistic regression

C Discriminative model like Support Vector Machine (SVM)

D Generative model like a Variational Autoencoder (VAE)

💡 Hint

Think about which model type can create new data samples.

Generative models like VAEs learn to create new data similar to training data, making them suitable for image generation.

❓ Metrics

advanced

Evaluating generative model quality

Which metric is commonly used to evaluate the quality of images generated by a generative model?

A Accuracy

B Inception Score (IS)

C Mean Squared Error (MSE)

D Precision

💡 Hint

Consider metrics designed specifically for generated images.

Inception Score measures how realistic and diverse generated images are, making it popular for generative model evaluation.

🔧 Debug

advanced

Debugging a discriminative model training issue

Given this code snippet training a discriminative model, what is the likely cause of the error?

model = LogisticRegression()
model.fit(X_train, y_train)
preds = model.predict(X_test)
print(f"Accuracy: {accuracy_score(y_test, preds)}")

Error message: ValueError: Found input variables with inconsistent numbers of samples

A X_train and y_train have different numbers of samples

B The model is not suitable for classification

C accuracy_score is not imported

D X_test contains missing values

💡 Hint

Check the shapes of training data and labels.

The error indicates that the features and labels arrays do not have matching sample counts, which is required for training.

❓ Predict Output

expert

Output of a simple generative model code

What is the output of this Python code simulating a simple generative model sampling from a Gaussian distribution?

import numpy as np
np.random.seed(42)
mean = 0
std_dev = 1
samples = np.random.normal(mean, std_dev, 3)
print([round(s, 2) for s in samples])

A [0.5, -0.14, 0.65]

B [0.5, -0.14, 0.45]

C [0.5, -0.14, 0.54]

D [0.5, -0.15, 0.65]

💡 Hint

Run the code or recall numpy's normal distribution output with seed 42.

With seed 42, np.random.normal generates approximately [0.4967, -0.1383, 0.6477], rounded to [0.5, -0.14, 0.65].