Quiz on Machine Learning Classification

1. What are two examples of classification? Select two answers.

A. Determine when a heater is on or off based on temperature data
Correct. The classifier distinguishes between on or off with temperature and temperature derivatives as the features.
B. Translate the numbers or letters from a handwritten message to ASCII text
Correct. The classifier analyzes the pixels of each letter to determine the alpha-numeric value.
C. Develop a mathematical relationship between heater level (0-100%) and temperature (20-70°C)
Incorrect. This is an example of regression where the label (output) is not a discrete outcome but is a continuous value
D. Safety override that automatically shuts off equipment when it is too hot
Incorrect. A safety override is an example of an interlock condition that protects equipment.

2. What answers are correct for supervised learning? Select all that apply.

A. Requires labeled data that reveals the measured or true outcome
Correct. Unsupervised learning does not use labeled data that are measured outputs. All of the other answers are also correct.
B. Data are divided into training and test sets
Correct. Test sets and validation are important to confirm that the training is effective. All of the other answers are also correct.
C. Supervised learning methods include Logistic Regression, Support Vector Classifier, and Random Forrest
Correct. These are all supervised learning methods. All of the other answers are also correct.
D. Supervised learning methods include Artificial Neural Network, K-Nearest Neighbors, and Stochastic Gradient Descent
Correct. These are all supervised learning methods. All of the other answers are also correct.

3. What method can be used with data that has a high fraction of missing labels (reported outputs)? Select two correct answers.

A. Supervised learning
Incorrect. Supervised learning requires labels for all feature instances
B. Semi-supervised learning
Correct. Semi-supervised learning is training with some labels that are missing
C. Unsupervised learning
Correct. Unsupervised learning does not use labels but may require input such as number of components to build a classifier
D. Reinforcement learning
Incorrect. Iteratively learning with reinforcement learning from data requires labels as a measure of the reward. In this particular case, reinforcement learning is not needed because the data is in one batch and does not appear sequentially over time.