Lecture 10 - Machine Learning in Affective Computing

Question 1

Difference between psychophysiology and affective computing:

Answer 1

Psyphy

• Gain knowledge
• Averages over groups of people
• With a considerable delay (at the end of an experiment for example)

AC

• Make applications
• Focused on individual
• Update in real time, directly when measured

Question 2

Standard error in mean gives …

Variance in measurement gives …

Answer 2

1. Significance

2. Effect size

Question 3

How does machine learning work in AC?

Answer 3

Combine parameters such as #ns-SCRs and SCL, and try to fit a line dividing two different zones (or more of course)

Question 4

Machine learning in general has more of everything:

Answer 4

2 or more states
1 or more signals
1 or more features per signal (2 or more in total)

Question 5

Steps of developing the classifier:

Answer 5

Training set:

• used for developing the classifier
• consisting of N features
• give labels for all states to be classified (ground truths)

Validation set:

• used for testing the classifier
• yields the accuracy of the model