GitHub - fzmath/diff-spectral-element: Differentiable spectral element method (DSEM) for PDE-constrained optimization. Embeds GLL spectral element solvers into PyTorch autograd, replacing hand-derived adjoints with automatic differentiation. Matrix-free CG, 1D/2D/3D, L-shaped domains, nonlinear Burgers.

Name		Name	Last commit message	Last commit date
run_3d.py		run_3d.py
spectral_element.py		spectral_element.py
spectral_element_2d.py		spectral_element_2d.py
spectral_element_3d.py		spectral_element_3d.py
test_2d.py		test_2d.py
test_2d_verify.py		test_2d_verify.py
test_adjoint_verify.py		test_adjoint_verify.py
test_alpha_sensitivity.py		test_alpha_sensitivity.py
test_burgers_nu.py		test_burgers_nu.py
test_burgers_nu2.py		test_burgers_nu2.py
test_cg_backward.py		test_cg_backward.py
test_cg_backward2.py		test_cg_backward2.py
test_elasticity.py		test_elasticity.py
test_fwd_bwd_timing.py		test_fwd_bwd_timing.py
test_grad_convergence.py		test_grad_convergence.py
test_gradient_opt.py		test_gradient_opt.py
test_lshape.py		test_lshape.py
test_lshape_revised.py		test_lshape_revised.py
test_multiseed.py		test_multiseed.py
test_reviewer_fixes.py		test_reviewer_fixes.py
test_scalability.py		test_scalability.py
test_sem.py		test_sem.py

About

Differentiable spectral element method (DSEM) for PDE-constrained optimization. Embeds GLL spectral element solvers into PyTorch autograd, replacing hand-derived adjoints with automatic differentiation. Matrix-free CG, 1D/2D/3D, L-shaped domains, nonlinear Burgers.

adjoint-method automatic-differentiation matrix-free numerical-analysis pde-constrained-optimization pytorch spectral-elements spectral-methods

Activity

0 stars

0 watching

0 forks

Report repository