I'm working towards graduate research in trustworthy AI and ML security, reproducing papers I want to understand, and writing up what I learn along the way. I want AI systems that are more transparent and evaluable, and I want to be part of the work that gets us there.
If you're working on trustworthy ML, I'd love to hear from you! Always looking to collaborate on interesting projects. :)
Recent writing
All posts →- Recovering confidence from labels alone 3/6
Part 2 got us a baseline. Now we ask the model about wobbled copies of a point, and use how reliably it still gets them right as a stand-in for confidence. It beats the baseline, though the obvious way of counting the answers does worse than not augmenting at all.
- The gap attack, and some ML fundamentals 2/6
To build the gap attack I first had to learn how to train a neural network, so this post is mostly the ML setup (CNNs, tensors, loss, gradient descent), with the working attack at the end.
- What is a membership inference attack? 1/6
The first in a series cataloguing my journey reproducing papers I want to properly understand. Starting with the core idea and vocabulary of membership inference attacks.