Introduction to machine learning

Question 1

What is an AI?

Answer 1

Algorithm capable of learning and making decisions

Question 2

What is the machine learning pipeline?

Answer 2

- Training:
     Load data
     Extract features
     Train model
     Evaluate
- Testing:
    Load data
    Extract features
    Predict using model
    Evaluate

(Only difference is that you initally train, and then use the trained model when testing)

Question 3

Briefly describe the 4 types of machine learning

Answer 3

• Supervised: Humans “teach” the algorithm
• Unsupervised: Algorithm does everything by itself
• Semi-supervised: In between the two previous
• Reinforcement: Algorithm left to itself, but gets
  rewarded if it does something correctly

Question 4

What is classification? How do you calculate loss in classification?

Answer 4

Classification is a classifier used in machine learning. The output is discrete. Loss i calculated by counting the number of misclassified samples.

Question 5

What is regression? How do you calculate loss in regression?

Answer 5

Regression is another classifier. The output is continuous. Loss is calculated by squaring (to avoid outliers) and adding up and taking the average.

Question 6

What are some other classifiers (besides classification and regression)?

Answer 6

Mainly tree based models, like Random Forest. Works by using a combination of N decision trees.

Question 7

What is bagging?

Answer 7

Combining several classifiers. Can reduce variance.

Question 8

What is the difference between classification and regression? Explain with an example if possible.

Answer 8

Classification has a discrete (categorical). If you are trying to predict house prices, it can tell you which houses might sell for below or above the predicted price (categories).

Regression has a continuous output (numerical). So if predicting housing prices, it actually gives you the predicted price.