Training data shapes everything
A model only knows what its training data showed it. If a fruit-sorting model saw only red apples, a green apple may confuse it. The data is the model’s entire experience of the world.
This is why data quality matters more than clever code. Missing groups, mislabelled examples and old information all end up inside the model’s behaviour.
After training, the model is tested on examples it has never seen. Doing well on fresh examples is the real test, because memorising the training set is not the same as learning.
Why is a model tested on examples it has not seen before?
Choose the answer you think is right.