Pre-built implicit layer architectures with O(1) backprop, GPUs, and stiff+non-stiff DE solvers, demonstrating scientific machine learning (SciML) and physics-informed machine learning methods

Efficient and Scalable Physics-Informed Deep Learning and Scientific Machine Learning on top of Tensorflow for multi-worker distributed computing

Public repository for the proposal “Physics-Informed Machine Learning Simulator for Wildfire Propagation” - MLJC University of Turin - ProjectX2020 Competition (UofT AI)

Code for "Learning Local Control Barrier Functions for Safety Control of Hybrid Systems"

Comparison of numerical solutions of the 1-D time-independent Schrödinger equation obtained through FDM, FEM and the neural network approach.

Numerical analysis of the convergence behavior of adjoint-optimized neural PDEs

Neural PDEs tested on protein networks

Multi-Architecture Coupled Ensemble Physics-Informed Neural Networks (MACE-PINN) for solving coupled partial differential equations. Implements parallel subnetworks with Fourier embeddings and adaptive loss weighting for Gray-Scott and Ginzburg-Landau reaction-diffusion systems.

Offline Reinforcement Learning with neural PDEs.

Numerical solution and uncertainty quantification of Pennes' bioheat transfer equation in 1-D using deep neural network solver.