Skip to main content
Workshop exercise
In Progress

Facing fairness in facial recognition practical

In this exercise, you will compute tone weights, build per-sample weights, fine-tune the final epochs with sample_weight, and recompute weighted F1 and max-FNR gap.

Instruction icon

Challenge instructions and resources

Download the Jupyter notebook bundle for the practical activity.

Success criteria​

  • Compute tone weights.
  • Build per-sample weights.
  • Fine-tune the last three epochs with sample_weight.
  • Recompute weighted F1 and max-FNR gap, aiming for gap ≤ 0.05 while keeping F1 at or above baseline.

Activity flow​

  • Work through the notebook instructions.
  • Compare results with peers in the breakout room.
  • Summarise what changed, what improved, and which trade-offs remained.