Guide-yakabatana mibvunzo · Pakati Nhanho

Bias Mitigation Techniques: Pre-, In- and Post-Processing Quiz

Classify pre-, in-, and post-processing methods and evaluate their metric tradeoffs, assumptions, and limitations in real workflows.

Nzira dzekutungamira dzinoenderanaBias Mitigation Pre In And Post Processing
Mubvunzo 1 ye 8

Ndeipi kupindira kuri pre-processing?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Ndeipi kupindira kuri pre-processing?
    • Changing the threshold after prediction
    • Adding a penalty to the learning objective
    • Changing training weights before fitting the model
    • Writing a public audit summary
  2. Which intervention is in-processing?
    • Adding a fairness constraint to the learning algorithm
    • Removing duplicate rows
    • Adjusting final scores after prediction
    • Changing the user notice
  3. Ndeipi kupindira iri post-processing?
    • Reweighting the training records
    • Changing annotation instructions before collection
    • Changing the model’s loss function
    • Adjusting decision thresholds after model training
  4. What does AIF360 Reweighing do?
    • Replaces every label with a majority vote
    • Assigns different weights to training instances
    • Guarantees equal accuracy
    • Removes the need for an audit
  5. Chii chinodiwa neFairlearn ThresholdOptimizer kuti ushandise fungidziro yeboka?
    • Sensitive-feature information and a selected constraint/objective
    • Only a model card
    • The model source code
    • A New York City license
  6. Sei nzira ichigona kuvandudza metric imwe chete asi ichiwedzera imwe?
    • Metrics always move together
    • The tool contains a coding error
    • Fairness definitions measure different properties and can involve tradeoffs
    • Accuracy is the only relevant measure
  7. Does reducing a measured disparity automatically make deployment lawful?
    • Yes, all toolkits certify legality
    • No; toolkit outputs do not decide whether the use is valid or lawful
    • Only when the metric is demographic parity
    • Yes, if accuracy is high
  8. What should be compared when selecting a mitigation method?
    • Fairness metric alone
    • Only overall accuracy
    • The library’s popularity
    • A baseline and candidates on held-out data across groups, with uncertainty and tradeoffs