Tackling Information Leakage in Concept Bottleneck Models
Deep Neural Networks (DNNs) are powerful black boxes, but their lack of transparency makes them unsuitable for high-stakes domains like healthcare. Concept Bottleneck Models (CBMs) …
Selected Projects
I enjoy making things. Here are a selection of projects that I have worked on over the years.
Adaptive Random Walk Metropolis with Spectral Preconditioning
High-Level Overview In Bayesian inference, we often need to sample from complex, high-dimensional posterior distributions that are non-conjugate. While MCMC algorithms like the …
Toward using GANs in astrophysical Monte-Carlo simulations
High-Level Overview Accurate modeling of spectra from X-ray sources requires computationally expensive Monte-Carlo (MC) simulations. These simulations must sample from numerous …