2026 · Journal of Computational Physics
Binned Spectral Power Loss
Frequency-domain loss design to reduce spectral bias in long-horizon forecasting of chaotic systems.
Open project details
Dibyajyoti Chakraborty
I am a PhD student in Informatics at Penn State University. My research focuses on deep learning for dynamical systems, with an emphasis on theoretical advances for forecasting high-dimensional chaotic systems and applications in turbulence and weather forecasting.
Research Works
Selected projects
2026 · Machine Learning: Earth
Multimodal Atmospheric Super-Resolution
Score-based diffusion for reconstructing high-resolution atmospheric states from low-fidelity multimodal observations.
Open project details
2026 · arXiv
Adaptive Diffusion Posterior Sampling
Probabilistic long-horizon turbulent-flow forecasting with adaptive sensor placement and diffusion posterior sampling.
Open project details
2025 · Physical Review Fluids
Differentiable Turbulence
End-to-end differentiable LES for learning generalizable SGS closures in two-dimensional turbulence.
Open project details
2025 · arXiv
MoWE: A Mixture of Weather Experts
A Vision Transformer gating network that combines existing weather models into a stronger forecast.
Open project details
2024 · CMAME
Divide and Conquer for Chaotic Dynamics
Trajectory-windowed neural ODE training with discontinuity penalties for chaotic systems.
Open project details
