About me

I am a Postdoctoral Researcher at Ecole Polytechnique Federale de Lausanne (EPFL) since November 2020.

My research focuses on reliable machine learning and the design and study of expressive models that are robust to noise and generalize well in out-of-distribution data. Concretely,
  • I have introduced polynomial networks (PNs) that capture high-degree interactions between inputs. My short-term goals are to understand the inductive bias and properties of existing architectures through empirical and theoretical studies. I am interested in the complete theoretical understanding of (neural/polynomial) networks, including their expressivity, trainability, generalization properties, and inductive biases. My recent work has provided the first characterization of the generalization of this class of functions, and I have focused on the spectral bias of high-degree polynomials, highlighting how PNs can learn higher frequency functions faster than regular feed-forward networks.
  • My goal is to understand the extrapolation properties of existing networks and make improvements to their performance, especially in the context of conditional generative models. In the short-term, I will continue to explore the robustness of these models to malicious attacks, as well as the impact of adversarial perturbations on different classes. In the long-term, I plan to design models that are both robust and fair, and can generalize well to unseen attribute combinations.