Real World Maths Busters

The Maths Behind Everyday Life

The story, the formula and a worked example — see exactly where this maths shows up in the world around you.

Back to all stories
Technology 3 min read

When AI Memorises Instead of Learning

Overfitting: 99% accuracy on training data, 60% on new data.

An overfitted model is like a student who memorises answers but can’t solve new problems. The gap between training and test accuracy reveals overfitting.

The Generalisation Gap

Training accuracy = 99%, test accuracy = 60%. The gap is 39 percentage points — a clear sign of overfitting. A well-generalised model might show 95% training and 93% test.

Measuring overfitting

  1. 1

    Overfitted model

    99%−60%=39% gap99\% - 60\% = 39\% \text{ gap}
  2. 2

    Good model

    95%−93%=2% gap95\% - 93\% = 2\% \text{ gap}
Answer

A large gap means the model memorised rather than learned

The takeaway

The gap between training and test accuracy is the thermometer for overfitting — smaller is better.

Back to all stories