Learning in feed-forward neural network theory is a complex information study that seeks to reconstruct a desired input-output function from contributed examples (i.e., functions). This research utilizes recent advances in radial basis function artificial neural network (RANN) learning to estimate regime dependent spillover elasticity metrics between the municipal bond returns of the US and the South African (SA) government 10-year bond.  We use a unique trade-level database to examine municipal bond market volatility across SALT-impacted states and the South African government bond market.  A two-regime learning simulation is tested.  The first regime covers the Tax Cuts and Jobs Act of 2017 (i.e., TCJA).  The second regime focuses on the time-frame defined by the global recognition of the COVID-19 virus. Elasticity estimates from the first regime uncovered spillover caused by variation in the export of New Hampshire chickens to SA. Results from models of the second regime find a shift in spillover effects as well as a change in the elasticity of sovereign idiosyncratic market risk.  The RANN information complexity algorithm deployed in this research proved capable of accounting for the shape and altered transmission of bond market volatility spillover between US states and SA.