_index.org

simulated annealing

Last edited: August 8, 2025

Uses a randomness temperature, which starts high for exploration and slowly decreases by one of the Simulated Annealing Schedules. Whenever you are about to take a step, each step must be “accepted” with probability given by the Metropolis Criteria.

Intuition: “sometimes, you try something that’s bad, to jump out of local optima.”

Metropolis Criteria

\begin{align} P(\text{accept}) = \begin{cases} 1, \text{if}\ \Delta y \leq 0\\ \min \qty(e^{-\frac{\Delta y}{t}}, 1), \text{otherwise} \end{cases} \end{align}

where \(t\) is the temperature; meaning—-even if the next point \(\Delta y > 0\) (meaning its bad, because it increases loss), we sometimes take that ponit still.

single party control

Last edited: August 8, 2025

Single-Objective Optimization

Last edited: August 8, 2025

SIR Model

Last edited: August 8, 2025

The SIR Model is a model to show how diseases spread.

  • Susceptible – # of susceptible people
  • Infectious — # of infectious people
  • Removed — # of removed people

Compartmental SIR model

S => I => R [ => S]

So then, the question is: what is the transfer rate between populations between these compartments?

Parameters:

  • \(R_0\) “reproductive rate”: the number of people that one infectious person will infect over the duration of their entire infectious period, if the rest of the population is entirely susceptible (only appropriate for a short duration)
  • \(D\) “duration”: duration of the infectious period
  • \(N\) “number”: population size (fixed)

Transition I to R:

skinning

Last edited: August 8, 2025

skinning is a notion in graphics which solves Cloth Fitting Task which adds a prior \(S\qty(\theta)\) which is not learned to make the learning of \(D\qty(\theta)\) easier

\begin{equation} \phi = f_{}\qty(\theta) = S \qty(\theta) + D \qty(\theta) \end{equation}

so then, when we make a model, keeping around \(S\qty(\theta)\) which is learned by physical dynamics, and constraining a displacement map \(D\qty(\theta)\) within each square \(S\qty(\theta)\).

We can then rasterize the diffs into RGB space into a grid, then we can run a CNN.