See Free energy principle for context.
We can describe it using the Fokker-Planck equation. It's steady state solution:
p˙=0=∇⋅(Γ∇−f)p
∇⋅Γ∇p=p∇⋅f+f⋅∇p
p∇⋅Γ∇p=∇⋅f+f⋅p∇p
∇⋅Γ∇lnp−Γ∇p⋅∇(p1)=∇⋅f+f⋅∇lnp
∇⋅Γ∇lnp+p21Γ∇p⋅∇p=∇⋅f+f⋅∇lnp
∇⋅Γ∇lnp+Γ∇lnp⋅∇lnp=∇⋅f+f⋅∇lnp
We can decompose any vector field into an irrotational (curl-free), and solenoidal (divergence-free) component (Helmholtz decomposition), which can be expressed as the so-called standard form:
f=(Γ+Q)∇V
where Q is antisymmetric, and Γ is symmetric. Substituting in above equation
∇⋅Γ∇lnp+Γ∇lnp⋅∇lnp=∇⋅Γ∇V+(Γ+Q)∇V⋅∇lnp
Notice that (Γ+Q)∇V⋅∇lnp=∇VT(Γ+Q)∇lnp=∇VTΓ∇lnp because Q is antisymmetric. Using this, if we assume that V=lnp, then the above equation for stationarity is satisfied.
Proof from paper: see Free_energy_principle_lemmaD1a.png and Free_energy_principle_lemmaD1b.png
It is straight-forward but fundamental result means that the flow of any ergodic random dynamical system can be expressed in terms of orthogonal curl- and divergence-free components, where the (dissipative) curl-free part increases value while the (con-servative) divergence-free part follows isoprobability con-tours and does not change value. Crucially, under this decomposition value is simply negative surprise: lnp(x∣m)=V(x)=−L(x∣m). It is easy to show that surprise (or value) is a Lyapunov function for the policy
V˙=∇V⋅f=∇V⋅Γ⋅∇V+∇V⋅∇×W=∇V⋅Γ⋅∇V≥0