Engineering4 min read
Calibrate a scientific model with Gauss–Newton
Recover amplitude, decay rate, and offset in a nonlinear least-squares inverse problem using a verified residual Jacobian and functional Gauss–Newton steps.
DOctober 5, 2026
ReadBlog
Notes from the derivative engine
Engineering deep-dives, research, and benchmarks you can re-run. Written by the team building Derivon.
Tags:
Engineering4 min read
Train differentiable regions and local experts
Learn a soft split and two regional regressors together, validate held-out predictions, and measure what changes when routing is hardened.
DOctober 5, 2026
ReadEngineering4 min read
A Poisson PINN without a full Hessian
Train a two-dimensional Poisson PINN using direct Laplacian contraction, then inspect solution error, boundary error, and residual error separately.
DOctober 5, 2026
ReadEngineering4 min read
A fourth-order beam PINN
Solve a clamped Euler–Bernoulli beam with a nonuniform load, forward Taylor jets, exact boundary conditions, and independent solution checks.
DOctober 5, 2026
ReadEngineering2 min read
Structure belongs in the model contract
A hard boundary construction and a soft residual penalty provide different guarantees.
DOctober 4, 2026
ReadResearch2 min read
VMC integrations and the public substrate boundary
Separate derivative primitives from the wavefunction and local-energy contracts of an application.
DOctober 4, 2026
Read