Skip to content

ai_ml_basics_quiz.md

Latest commit

 

History

History
215 lines (142 loc) · 7.66 KB

ai_ml_basics_quiz.md

File metadata and controls

215 lines (142 loc) · 7.66 KB

AI & Machine Learning Basics Quiz

Q1. What does “AI” stand for?

  • Artificial Intelligence
  • Automated Integration
  • Applied Innovation
  • Algorithmic Index

Explanation: AI refers to machines designed to mimic human intelligence.

Q2. Which of the following is an example of supervised learning?

  • K-Means Clustering
  • Linear Regression
  • Principal Component Analysis
  • DBSCAN

Explanation: Linear regression is a supervised learning algorithm because it learns from labeled data.

Q3. What is overfitting in machine learning?

  • When the model is too simple
  • When the model performs well on training data but poorly on new data
  • When the data is corrupted
  • When the algorithm runs too slowly

Explanation: Overfitting occurs when the model learns noise from the training data instead of general patterns.

Q4. Which of these is a common activation function in neural networks?

  • Mean Squared Error
  • Gradient Descent
  • ReLU
  • Random Forest

Explanation: ReLU (Rectified Linear Unit) is widely used in neural networks as an activation function.

Q5. What is the difference between AI and ML?

  • ML is a subset of AI focused on learning from data
  • AI is a subset of ML
  • AI and ML are the same
  • ML refers to hardware, AI to software

Reference: Intro to AI & ML

I'll analyze the existing quiz and provide two additional questions while maintaining the same format:

Q6. What is Deep Learning?

  • A subset of machine learning using neural networks with multiple layers
  • A type of database management system
  • A programming language for AI
  • A method of storing large datasets

Explanation: Deep Learning is a specialized subset of machine learning that uses neural networks with multiple (deep) layers to progressively extract higher-level features from raw input. It's particularly powerful for complex tasks like image and speech recognition.

Source: Deep Learning - MIT Press

Q7. Which of these is NOT a common type of machine learning?

  • Supervised Learning
  • Unsupervised Learning
  • Reinforcement Learning
  • Perpetual Learning

Explanation: While Supervised, Unsupervised, and Reinforcement Learning are the three main types of machine learning, "Perpetual Learning" is not a standard classification. The main categories are:

  • Supervised Learning (learns from labeled data)
  • Unsupervised Learning (finds patterns in unlabeled data)
  • Reinforcement Learning (learns through trial and error with rewards/penalties)

Source: Types of Machine Learning - Stanford University CS229

Q8. What is the main goal of reinforcement learning?

  • To learn from labeled data
  • To reduce dimensionality of data
  • To learn optimal actions through rewards and penalties
  • To predict continuous values

Explanation: RL agents learn to take actions that maximize cumulative reward through feedback loops.

Reference: Reinforcement Learning - Sutton & Barto

Q9. Which algorithm is commonly used for classification tasks?

  • Decision Tree
  • K-Means
  • Apriori
  • DBSCAN

Explanation: Decision Trees are popular for classification because they split data based on feature values to predict labels.

Reference: Decision Tree Classifier - Scikit-learn

Q10. What is the purpose of “gradient descent” in training neural networks?

  • To increase the loss function
  • To minimize the loss function
  • To normalize input data
  • To calculate feature importance

Explanation: Gradient Descent minimizes the loss by updating model parameters opposite to the gradient direction.

Reference: Gradient Descent Explained - Coursera

Q11. In which situation would you most likely use clustering algorithms?

  • When you have labeled data
  • When you want to group similar data points without labels
  • When you want to perform regression analysis
  • When predicting future prices

Explanation: Clustering is an unsupervised learning method used to group unlabeled data by similarity.

Q12. What does a confusion matrix show?

  • Model loss over time
  • True positives, false positives, true negatives, and false negatives
  • Feature correlations
  • Decision boundaries

Explanation: A confusion matrix summarizes prediction outcomes for classification models.

Q13. What is the purpose of feature scaling?

  • To ensure all features contribute equally to model training
  • To remove irrelevant features
  • To encode categorical variables
  • To reduce overfitting

Explanation: Feature scaling standardizes numerical ranges, helping algorithms like gradient descent converge efficiently.

Q14. What does “bias-variance tradeoff” refer to?

  • The balance between model simplicity and its ability to generalize
  • The difference between input and output bias
  • The trade between accuracy and recall
  • The time-memory tradeoff

Explanation: Low bias can cause overfitting, while high bias leads to underfitting — good models find balance.

Q15. What metric is best for imbalanced classification problems?

  • Accuracy
  • MSE
  • F1-Score
  • R² Score

Explanation: F1-Score balances precision and recall, making it ideal for imbalanced datasets.

Q16. What is the main purpose of a validation set?

  • To train the model
  • To tune hyperparameters and evaluate performance
  • To test model performance on unseen data
  • To normalize the input features

Explanation: Validation sets help in model selection and hyperparameter tuning before final testing.

Q17. Which algorithm is best suited for dimensionality reduction?

  • KNN
  • Naive Bayes
  • PCA (Principal Component Analysis)
  • Random Forest

Explanation: PCA transforms high-dimensional data into fewer dimensions while retaining variance.

Q18. Which type of neural network is best for sequence data?

  • CNN
  • RNN
  • GAN
  • DBN

Explanation: RNNs (Recurrent Neural Networks) are designed for sequential inputs like text or time series.

Q19. What is the purpose of regularization in machine learning?

  • To prevent overfitting by adding penalty terms to the loss function
  • To normalize inputs
  • To improve data quality
  • To increase model complexity

Explanation: Regularization methods like L1/L2 reduce overfitting by constraining model weights.

Q20. Which of the following is NOT a type of ensemble method?

  • Random Forest
  • Gradient Boosting
  • Bagging
  • Dropout

Explanation: Dropout is a regularization technique in neural networks, not an ensemble method.

Q21. What is the main advantage of using cross-validation?

  • Provides a more reliable estimate of model performance
  • Reduces the need for a validation set
  • Speeds up training
  • Simplifies the model

Explanation: Cross-validation uses multiple folds to reduce variance in performance estimates.

Q22. Which statement best describes unsupervised learning?

  • Algorithms find patterns in data without predefined labels
  • Algorithms require labeled input
  • Models are trained with rewards and penalties
  • It’s only used for classification problems

Explanation: Unsupervised learning discovers hidden structures in unlabeled data.