blackjax.smc.solver

blackjax.smc.solver#

All things solving for adaptive tempering.

Functions#

dichotomy(fun, min_delta, max_delta[, eps, max_iter])

Solves for delta by dichotomy.

Module Contents#

dichotomy(fun, min_delta, max_delta, eps=0.0001, max_iter=100)[source]#

Solves for delta by dichotomy.

If max_delta is such that fun(max_delta) > 0, then we assume that max_delta can be used as an increment in the tempering.

Parameters:

  • fun (Callable) – The decreasing function to solve, we must have fun(min_delta) > 0, fun(max_delta) < 0

  • min_delta (float) – Starting point of the interval search

  • max_delta (float) – End point of the interval search

  • eps (float) – Tolerance for \(|f(a) - f(b)|\)

  • max_iter (int) – Maximum of iterations in the dichotomy search

Returns:

delta – The root of fun

Return type:

Array, shape (,)