Research
Last Updated July 2025
My research interests are in Differential equations and Machine Learning. Broadly, my research interests are using machine learning to study Differential equations while ensuring both physical interpretability and numerical validity. State estimation problems, where only partial data is available, and developing ML approaches to learn dynamical systems from incomplete information. Additionally, I am interested in the theory of deep learning i.e generalization theory of physics-informed machine learning models.
I am working on Port-Hamiltonian systems and neural network-based methods to model and learn them. Building on the Port-Hamiltonian Neural Network (PHNN) framework, I proved a universal approximation theorem for PHNNs, showing that they can approximate any dynamical system within this class. This result naturally extends to other variants, including compositional PHNNs and Stable Port-Hamiltonian Neural Networks (sPHNNs). Currently, I am focused on developing training methods that ensure these learned neural networks preserve stability.
Undergraduate Research
I worked under Dr. Hassan Safouhi to develop algorithms to efficiently compute tail probabilites using the $G_n^{(1)}$ transformation and the Slevinsky - Safouhi Formula6. The new algorithm reduces the complexity from the previous algorithm from $\mathcal{O}(n^4)$ to $\mathcal{O}(n)$.
Paper in progress.
Gaudreau, Philippe, Richard M. Slevinsky, and Hassan Safouhi. "Computation of tail probabilities via extrapolation methods and connection with rational and padé approximants". SIAM Journal on Scientific Computing 34.1 (2012): B65-B85. ↩
I was an intern at the Max Planck Institute for Dynamical Systems. My research consisted of extracting feature weights and maps from various models such as Convolutional Neural Networks, Convolutional Autoencoders (CAEs) and Individual CAEs (iCAEs)1 for the Discretized Incompressible Navier - Stokes Equations within a single and double cylinder and analyzing them.
For my research at the University of Alberta, Prof. Yaozhong Hu and myself have been thoroughly studying and analyzing Clair et al.'s work2 Determinantal Point Processes (DPP) for Image Processing. We aim to understand the paper better and extract valuable insights from its content and findings.
[1] Kim, Yongho, and Jan Heiland. "Convolutional Autoencoders, Clustering, and POD for Low-dimensional Parametrization of Navier-Stokes Equations". Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany, and Department of Mathematics, Otto von Guericke University Magdeburg, Magdeburg, Germany. 2023. ↩
[2] Launay, Agnès Desolneux, Bruno Galerne. "Determinantal Point Processes for Image Processing". SIAM Journal on Imaging Sciences, 2020, 14 (1), pp.304-348. ff10.1137/20M1327306ff. ffhal-02611259v2 ↩
During the Winter Semester of my third year, I did my undergraduate thesis under Prof. Bin Han. The title of my thesis was Linear Independence and Stability of Functions. I started off by studying preliminary material like Distribution Theory and Functional Analysis by going through Loukas Grafakos1 Classical Fourier Analysis. Then, we proved some results which included Linear independence of compactly supported distributions and relating the Kernel of the space of semi-discrete convolutions of a distribution to the Fourier transform of the distribution.
My thesis writing has been enlightening, providing me with valuable knowledge and skills. Despite the significant time commitment, I efficiently developed my paper writing skills through guidance and feedback from my supervisor.
We also had to do a write-up on a famous classical result in our field so I chose to do it on the Poisson Summation Formula. This improved my ability to summarize complex concepts and communicate them effectively to others.
The full thesis, including the poster, presentation and the clasical result can be accessed here [Thesis], [Poster], [Presentation], [Write-up].
Grafakos, Loukas. Classical and Modern Fourier Analysis. 3rd ed., Springer, 2014. ↩
Haven taken the Honors Ordinary Differential Equations (ODEs) course in the Winter '22 semester, taught by Prof. Mohammad Ali Niksirat, we were given the opportunity to work on a final project. I decided to work on proving the existence of Periodic Solutions for Second - Order Non - Autonomous Differential Equations.
Following this semester, I decided to work under Prof. Niksirat on the same project over the summer. We approached it using Degree Theory. I went through the first two chapters of Louis Nirenberg’s4 Topics in Non-Linear Functional Analysis. I reproved some of the theorems in the text for example the Brouwer's Fixed Point Theorem, which were the building blocks leading up to the main theorem which involved proving the existence of the solution.
The report, poster and proposal can be accessed here [Report], [Poster], [Proposal].
[4] Nirenberg, Louis. "Topics in Nonlinear Functional Analysis". New York University - Courant Institute of Mathematical Sciences, New York, NY. ↩
During the summer of 2020, after high school graduation, I discovered Professor Petra Bonfert-Taylor's course on Introduction to Complex Analysis at Wesleyan University on Coursera. This sparked my interest in Complex Analysis. In the following summer, I reached out to Prof. Manish Patnaik regarding research in the field of Analytic Number Theory, which involved studying the Moments of Quadratic Dirichlet L - Functions.
I went through the first five chapters of Michael Rosen’s1 Number Theory in Function Fields to get a decent understanding and the suitable background required to undertake the project. The penultimate proof involved in proving the functional equation for the Dirichlet L-series was so arduous that the idea of congregating around the problem took up most of my time. As this was my first research experience, it helped me in becoming more mathematically mature and also gave me a good insight into how research is usually done.
The report and proposal can be accessed here [Report], [Proposal].
Michael, Rosen. "Number Theory in Function Fields". Springer. 2002. ↩