The equity add-on is built in three layers. Risk positions are first aggregated at the reference entity level to produce an Effective Notional. Each reference entity then produces an Entity Addon equal to its Effective Notional multiplied by the supervisory factor.

The credit add-on is built in three layers. First, risk positions are aggregated at the reference entity level to produce an Effective Notional. Second, each reference entity produces an Entity Addon equal to its Effective Notional. Third, all Entity Addons within a hedging set are combined using a formula that separates risk into a systemic component — capturing net directional exposure to a common credit factor — and an idiosyncratic component

In this article, I describe my experience of working with AI on various aspects of regulatory capital and how one formula triggered the "Pattern Recognition".