Machine Learning

Reading notes on molecular-orbital-based machine learning, which predicts post-Hartree-Fock correlation energies from Hartree-Fock orbital features.

Reading notes mapping the machine learning force field toolkit of Unke et al. onto familiar quantum chemistry concepts.

From nonlinear features to kernels, and the trick that makes an infinite-dimensional feature space cost almost nothing.

Linear regression in primal and dual form, and least squares as the maximum likelihood answer.