Model Training & Adaptation
Spectral clustering
Grouping examples using patterns in how strongly they are connected or similar to one another.
Example
A method groups points that are linked in a similarity graph.
Why people use it
It groups cases using the pattern of connections between them.
What you'll hear
“Which points are connected through strong similarities?”
What this means for you
Inspect the similarity construction because it strongly shapes the groups.
Can you control it?
Developer-only
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Can the way connections are chosen affect the groups?
- Yes. Changing which examples count as related can change the groups the method finds.
- Does the choice of connections matter?
- Yes. A different way of defining similarity can produce different groups.
- Can it become expensive for a large collection?
- Yes. Building and analyzing the relationships can require substantial computing resources.