Technological Knowledge MCQs with Answers and Explanations

What is a generative adversarial network (GAN) primarily used for?

A. Classification
B. Clustering
C. Generating synthetic data
D. Dimensionality reduction

Explanation:
A GAN consists of a generator and a discriminator; it is used to generate new data instances similar to the training set.

In a neural network, why is it important to normalize input features?

A. To make the network more complex
B. To improve convergence speed and training stability
C. To prevent overfitting entirely
D. Normalization is not needed

Explanation:
Feature normalization ensures they have similar scales, which helps gradient descent converge faster.

What is the primary advantage of using a GPU over a CPU for training deep neural networks?

A. Faster single-thread performance
B. Large memory capacity
C. Parallel processing of many operations
D. Built-in neural network instructions

Explanation:
GPUs can perform many operations in parallel, greatly speeding up matrix computations in neural network training.

Which of the following tasks is an example of unsupervised learning?

A. Predicting housing prices
B. Grouping customers by purchasing behavior
C. Classifying emails as spam or not
D. Translating text from English to French

Explanation:
Clustering customers by behavioral similarity is an unsupervised learning task, since no labels are provided.

What does the term "bagging" refer to in ensemble learning?

A. A bag is placed around the dataset
B. Training multiple models on different random subsets and aggregating results
C. Sequentially training models to correct errors (boosting)
D. A feature engineering technique

Explanation:
Bagging (Bootstrap Aggregating) trains multiple models on bootstrapped data samples and averages their predictions to reduce variance.

What role does the "loss function" serve in training a neural network?

A. Generate predictions
B. Aggregate features
C. Calculate error to guide learning
D. Normalize data

Explanation:
The loss function quantifies the difference between predictions and true values, guiding weight adjustments.

Which metric is typically used to evaluate the performance of a binary classification model?

A. Euclidean distance
B. Mean Squared Error
C. Accuracy or F1-score
D. AUC-ROC

Explanation:
Accuracy and F1-score are common evaluation metrics for classification models.

In natural language processing, which model is specifically designed to handle sequential data by maintaining a hidden state that carries information through steps?

A. Convolutional Neural Network (CNN)
B. Recurrent Neural Network (RNN)
C. Support Vector Machine (SVM)
D. Decision Tree

Explanation:
RNNs are designed for sequential data, with hidden states capturing information across sequence elements.

Which of the following best describes reinforcement learning?

A. Learning from labeled examples
B. Learning from unlabeled data via clustering
C. Learning through trial-and-error with rewards
D. A technique for anomaly detection

Explanation:
Reinforcement learning involves an agent learning to make decisions by receiving rewards or penalties from the environment.

What is the purpose of an epoch in machine learning training?

A. A type of activation function
B. A full pass through the entire training dataset
C. A regularization technique
D. A specific type of loss function

Explanation:
One epoch means the learning algorithm has processed the entire training dataset once.