Theory and Functional Design

The CIDER framework separates three questions that are often combined in a single analytic density functional:

• Which electronic information is exposed to the model?

• Which exact constraints are built into that representation and energy form?

• How is the remaining functional dependence learned and evaluated?

The pages in this section answer those questions, covering the CIDER functional form, Gaussian-process fitting, and numerical evaluation of features. The feature definitions themselves are collected in Electronic Features in CiderPress.

Contents

• The CIDER Framework
• Uniform Scaling
• From Exchange Models to Full XC
• Machine Learning Framework for CIDER Functionals
• Numerical Evaluation of NLDF Features
• Numerical Evaluation in PySCF and GPAW