Measuring How Good the Model Is

To train a model, we need to measure how good its result is.

Think about an exam:

1. each task has points
2. wrong answers lose points
3. all points together become a final grade

AI models work in the same way. For every task, the model gets a score called loss:

1. high loss → bad result
2. low loss → good result

During training, the goal is to reduce the loss over time.