Free Energy, Entropy, and Temperature

\(F = E - TS\).

  • where
    • \(F\). is the free energy
      • The nature tries to diminish the free energy.
      • Minimizing $F$ involves a trade-off between energy minimization ($E\downarrow$) and entropy maximization ($S\uparrow$).
    • \(E\). is the energy
      • the internal energy or potential
      • the tendency to reach a low-energy, highly-ordered state (e.g., specific data patterns).”
    • \(S\). is the entropy
      • The statistical measure of microstates
      • a state with higher entropy has a higher probability of occurrence due to the larger number of possible configurations.
    • \(T\). is the temperature
      • Acts as a scaling factor that controls the balance
      • high $T$ promotes exploration (Entropy-driven), while low $T$ promotes exploitation (Energy-driven).



Enjoy Reading This Article?

Here are some more articles you might like to read next:

  • (DM Reconst.) Ch.3 Score-Based Perspective - From EBMs to NCSN
  • Variational Inference with Normalizing Flows
  • Classifier-Free Diffusion Guidance (CFG)
  • Guiding a Diffusion Model with a Bad Version of Itself (Autoguidance)
  • (Presentation PDF) High Resolution Image Synthesis with Latent Diffusion Models (Latent Diffusion)