Implementing Kolmogorov-Arnold Networks as Dense GEMMs in JAX: A Neural QMC Experiment
Replacing B-splines with orthogonal polynomials and jnp.einsum in a Slater–Jastrow–backflow wavefunction for H2.
Written by
Industrial Chemist, Former Equipment Management & ML Researcher | Architect of FermiKAN
Industrial chemist and former equipment manager turned AI for Science researcher. Passionate about bridging theoretical physical chemistry and modern machine learning. Architect of FermiKAN (a TPU-friendly Neural QMC architecture). Constantly exploring how neuro-symbolic approaches and AI thought partners can push the boundaries of computational chemistry.
Responses
Markdown supported · ⌘↵ to post
No responses yet.