Updating priors

by: Kevin Broløs and Meera Machado
(Feyn version 3.0 or newer)

---

We can change the behaviour of how we sample Models from a QLattice by supplying prior probabilities for each input.

This can be used to inform a QLattice of our prior beliefs of the input variables. Inputs with higher prior values are more likely to appear in the sampled Models.

In short, a prior probability of an input x in the context of a QLattice denotes our prior belief of the importance of x in predicting the output.

If no priors are given, all inputs are equally likely to appear when sampling a Model.

Example

Here's an example of how to update the priors:

import feyn

priors = {"x1": 1., "x2": 0.99, "x3": 0.98, "x4": 0.97}

ql = feyn.QLattice()
ql.update_priors(priors, reset=True)

Calculating the priors

Refer to Estimating priors to see how the priors can be calculated based on the mutual information between each input variable and the output.

---

Parameters of update_priors

priors

A Dict object where the keys are the names of the input variables and the values are the relative weights associated to each input.

reset

Default: True

A boolean that determines whether the existing priors should be reset (True) or merged with the new priors (False) when updating the QLattice.