Model and Descriptor Workflows

This section describes how CiderPress models are constructed, inspected, and extended. A trainable model and its mapped evaluator serve different stages of that workflow.

A trainable MOLGP or MOLGP2 retains the reaction observations, covariances, noise models, and control-point fitting state. Mapping converts that object into MappedXC or MappedXC2, which contains the objects required for efficient DFT evaluation.

Contents

• Model Objects and Mapping
• Inspecting Densities and Descriptors
• Training and Mapping Workflow
• Extending CiderPress