Supervised
Unsupervised
Description:
You give the machine the right answer
How it works:
The machine learns the patterns in the data that caused the observed outcome to occur
Examples:
Housing prices (sale data is available), dating apps (user choice is observed)
Description:
There are no observable right answers so the machine creates it's own from scratch
How it works:
Generally the machine groups like things together
Examples:
Segmentation (which users are like other users), feature detection in an image
Classification
Regression
Clustering
Description:
The right answer is putting the observation into a category, which can even be a simple yes/no
How it works:
A form of supervised learning where the output is a likelihood of each category occuring
Examples:
Whether a house sold or not (or sold above market, pre-listing, etc.), whether two users liked each other
Description:
The output is a continuous number (say between 1 and 10,000)
How it works:
Given a set of inputs, the machine predicts what the observed value will be
Examples:
Housing prices (given sq. ft and neighborhood), total sales in June, attendance at an event
Description:
Which of my things are like others of my things?
How it works:
The machine groups your observations together into clusters of things that are like each other
Examples:
Grocery store shopper analysis (soccer moms vs. grandparents vs. college kids)
Another lightning talk
S'morebasborg chats!