inference
Running a trained model on new input to get an answer.
Inference evaluates a trained model's fixed function on new input; unlike training, it changes no parameters.
Examples
Each new one-second accelerometer window is fed to the gesture model on the ESP32, which answers idle, shake, circle or tap: that is inference.
Inference never changes the weights, so a gesture remote cannot "adapt to each user" on the board by itself.
See also