AI Foundations
Activation function
A calculation that changes the signal passed through a part of an AI network.
Example
A calculation changes signals inside an AI so it can learn more complex patterns.
Why people use it
It helps AI turn intermediate calculations into patterns useful for learning.
What you'll hear
“Which rule transforms the values at this stage?”
What this means for you
Judge its effect within the complete system and task.
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
- Why not just pass every signal through unchanged?
- Changing signals helps the network learn more complicated patterns than a chain of simple straight-line calculations could represent.
- Does one activation suit every part of a network?
- No. Different places in a network may need different kinds of transformation.
- Can an activation affect how easily the network learns?
- Yes. The way values change can strengthen or weaken the learning signals passed through the network.