exponential_accept_rel

pyoptex.doe.cost_optimal.codex.accept.exponential_accept_rel(m0, m1, T)[source]

Computes the accept probability as an exponential function of the ratio between the new and old metric, and the temperature.

  • Positive metrics: (m1/m0)**(1/T)

  • Negative metrics: (m0/m1)**(1/T)

Note

This requires the metric to have one distinct sign.

Parameters

m0float

The old metric

m1float

The new metric

Tfloat

The current temperature

Returns

alphafloat

The accept probability.