# Compact-Support Telegraph Latent Sampler

- ID: 2818
- Canonical URL: https://synthcore.org/idea/2818/compact-support-telegraph-latent-sampler
- API JSON: https://synthcore.org/api/idea/2818.json
- API Markdown: https://synthcore.org/api/idea/2818.md
- Verification status: unverified
- Source: [arXiv:2608.29226](https://arxiv.org/abs/2608.29226)
- Category: sampling
- Solves: stability, sample-efficiency, accuracy
- ML areas: diffusion-sampling, diffusion, vae
- Math tags: stochastic-processes, dynamical-systems, probability, statistical-mechanics
- Ratings: usefulness 6/10; difficulty 5/10; novelty 8/10

## Idea description

Use the OU process driven by multiple dichotomous noises as a bounded colored-noise module for latent-variable or diffusion sampling. Its stationary forcing is compactly supported for fixed amplitudes, while heterogeneous amplitudes and switching rates create controllable non-Gaussian structure before the large-K Gaussian limit.

## Mathematical statement

Let x(t) be a latent perturbation satisfying dx/dt = -gamma x + Omega(t), with Omega(t) = sum_k v_k s_k(t), where s_k are independent symmetric telegraph signals with switching rates lambda_k. The deterministic relaxation contracts initial conditions at rate gamma, and the stationary support satisfies |x| <= sum_k v_k/gamma. The stationary variance is sum_k v_k^2/[gamma(gamma+2 lambda_k)]. Increasing K while distributing amplitudes across channels produces a central-limit crossover, whereas a few dominant channels preserve non-Gaussian stationary structure.

## Key formulas

- x(t) = exp(-gamma t) x(0) + integral from 0 to t of exp[-gamma(t-s)] sum_k v_k s_k(s) ds.
- |x(t)| <= exp(-gamma t)|x(0)| + [1 - exp(-gamma t)] sum_k v_k/gamma, which converges to sum_k v_k/gamma.
- Var_stat(x) = sum_k v_k^2/[gamma(gamma + 2 lambda_k)].
- The Gaussian-crossover diagnostic is excess kurtosis kappa_4(x)/kappa_2(x)^2 -> 0 as K -> infinity when no single v_k dominates.

## Implementation notes

Insert the module between a neural encoder and decoder, or use it to replace Gaussian perturbations in a diffusion or latent-energy sampler. Given latent state z_n, compute the learned deterministic drift b_phi(z_n,t_n), update the telegraph force with u_(n+1) = exp(-gamma dt) u_n + [1 - exp(-gamma dt)] sum_k v_k s_(k,n)/gamma, and then use z_(n+1) = z_n + dt b_phi(z_n,t_n) + beta dt u_(n+1). Each sign flips independently with probability 1 - exp(-lambda_k dt). Keep gamma, lambda_k, and v_k fixed initially; tune only beta to match the variance of a Gaussian baseline. The exact relaxation and bounded-support equations come from the paper, while interaction with the learned drift must be evaluated empirically. First test on a two-dimensional mixture of Gaussians or checkerboard distribution using a small score network. Compare Euler-Maruyama Gaussian noise, standard colored OU noise, and telegraph-driven noise. Plot latent increments, radius distributions, mode coverage, and stability failures. With the learned drift disabled, the predicted stationary variance must match the formula within 10% and |u| must never exceed sum_k v_k/gamma. With one or two channels, histograms should show non-Gaussian edge structure; with many equal small channels, excess kurtosis should approach zero.

## Disclaimer

AI-generated research hypothesis, automatically tested. Not peer-reviewed.