Skip to content

MIT 6.S191 Lab 2: From MNIST to Facial Debiasing with a DB-VAE

Aug 22, 2026 1 min
TL;DR In the 2026 lab, part 1 classifies MNIST with dense and convolutional networks; Part 2 learns a facial latent distribution with a DB-VAE and changes training sampling.
Table of Contents
  1. Before you begin
  2. Recommended sequence
  3. Completion criteria
  4. Limits
  5. References

🌏 中文版

Lab 2 in the official MIT 6.S191 2026 repository is Lab 2: From MNIST to Facial Debiasing with a DB-VAE. Part 1 classifies MNIST with dense and convolutional networks; Part 2 learns a facial latent distribution with a DB-VAE and changes training sampling. This article pins the 2026 branch so later changes to master do not silently alter the exercise.

Before you begin

The official 2026 README specifies Google Colab, Python 3, and a GPU runtime. Copy the notebook to your Drive and run it from the beginning. Put API keys in the notebook's secret manager—never in a shareable cell or Git commit.

  1. Use Part 1 to verify convolution shapes and the training loop
  2. Record aggregate results and concrete failures from the baseline detector
  3. Compare the same cases after the DB-VAE without treating one metric as a fairness verdict

Solve one TODO at a time. Write the expected input and output shapes before executing the cell; when something fails, preserve the error and your reason for the fix. Public solutions are for final comparison, not initial copying.

Expected outputs include MNIST results, baseline face-detection results, and the same cases evaluated after DB-VAE resampling. Common failures include putting the image channel in the wrong dimension and reporting only aggregate accuracy without preserving failed face examples.

Completion criteria

Keep a notebook copy, one reproducible end-to-end run, and a short conclusion: what the model did correctly, where it failed, and which variable you would change next. A service-dashboard screenshot does not replace model outputs and an experiment record.

Limits

This is a course experiment, not a fairness fix proven across all populations. Data, measurement, and deployment context still require a separate audit.

References