IndisputableMonolith.Information.FEPBridgeFromJCost
The module supplies the local quadratic proxy for KL divergence in a single log-ratio coordinate, derived from J-cost. Researchers building the information bridge to variational free energy in Recognition Science cite these results when linking ledger costs to the KL term. The module consists of definitions for the proxy together with lemmas on its first and second derivatives that confirm agreement at equilibrium.
claimThe local quadratic proxy for KL divergence is the function $q(x)$ in the log-ratio coordinate $x$ satisfying $q(1)=0$, $q'(1)=0$, and $q''(1)$ matching the second derivative of the exact KL at equilibrium.
background
The module sits in the Information domain and imports the Cost module, which defines the J-cost function obeying the Recognition Composition Law. Its central construction is the quadratic proxy klQuadratic for KL divergence together with its derivatives. The downstream VariationalFreeEnergyFromRCL module uses this proxy inside the Friston functional $F[q;p]=\mathbb{E}q[E]+\mathrm{KL}[q||p\mathrm{prior}]$.
proof idea
This is a definition module, no proofs.
why it matters in Recognition Science
The module feeds the Information aggregator that assembles the information-theoretic foundation of Recognition Science and supplies the KL quadratic proxy required by VariationalFreeEnergyFromRCL. It thereby connects J-cost ledger coordinates to the free-energy descent property.
scope and limits
- Does not supply global error bounds away from equilibrium.
- Does not treat multi-dimensional or non-log-ratio coordinates.
- Does not derive the monotone descent of the free energy itself.
- Does not reference the phi-ladder or spatial dimension D=3.
used by (2)
depends on (1)
declarations in this module (16)
-
def
klQuadratic -
theorem
jcost_log_exact -
theorem
klQuadratic_zero -
theorem
hasDerivAt_klQuadratic -
theorem
deriv_klQuadratic_zero -
theorem
hasDerivAt_deriv_klQuadratic_zero -
theorem
hasDerivAt_deriv_Jlog_zero -
theorem
jcost_kl_same_second_order_at_equilibrium -
inductive
FEPStateClass -
abbrev
Coupling -
def
HasMarkovBlanketSparsity -
def
HasLedgerBoundarySparsity -
theorem
markov_blanket_sparsity_iff_ledger_boundary_sparsity -
structure
FEPBridgeLocalCert -
def
fepBridgeLocalCert -
theorem
fep_bridge_local_cert_holds