Topics in Information Theory and Coding (EE5848) : Aug-Nov 2026.

Differential Privacy for Statistics and Machine Learning

In this topics course we will be covering Differential Privacy for Statistics and Machine Learning this year. Tentative outline is given below.

• Instructor

  • Myna Vajha

• Timings and Venue

  • Slot R: Tuesday: 2.30-4pm, Friday: 4-5.30pm

  • Segments: 3-6

  • Credits: 2

  • Venue: EE-20F

  • First class on Tuesday, September 1, 2.30pm

• Outline (Tentative)

  • Indentification Attacks

  • Reconstruction Attacks

  • Differential Privacy

    • Randomized Response

    • Laplace Mechanism

    • Exponential Mechanism

    • Composition Theorems and Properties

  • Approximate Differential Privacy

    • Gaussian Mechanism

  • Applications to Statistics

    • Private mean estimation

    • Private Hypothesis Testing

  • Applications to Machine Learning

    • Private Emperical Risk Minimization

    • Private Stochastic Gradient Descent

• References

  • Cynthia Dwork and Aaron Roth: The Algorithmic Foundations of Differential Privacy

  • Salil Vadhan: The Complexity Of Differential Privacy

  • Lectures on Differential Privacy by Gowtham Kamath

  • Lectures on Differntial Privacy by Adam Smith and Jonathan Ullmann

  • Joseph P. Near and Chike Abuah: Programming Differential Privacy