| Training | Process of determining the model parameters (weights) that will allow an accurate mapping of the input to the resulting output (prediction) |
| i.e. so as to minimise the loss |
| (empirical risk minimisation) |
| Inference | Applying the already trained model to a set of inputs without known labels |
| Regression | Prediction of continuous values |
| s.
simple regression model |
| Classification | Prediction of discrete values (classes) |
| Loss | Difference between prediction and actual output |
| Penalty for a bad prediction |
| Hyperparameters | Configuration settings for tuning how the model is trained |