Quantum Rejection Sampling
Learn the concept of quantum rejection sampling.
We'll cover the following...
Rejection sampling is a straightforward method. We create samples and pick the ones that contain the evidence we’re interested in. The problem with this type of sampling is that we generate many samples that we don’t end up using. Let’s look at the following case of calculating the probability of female children being favored by a norm.
Evaluate the effectiveness
norm_mes = get_trained_qbn_norm(train, trained_params)print("{:.2f} of female children were favored by a norm ".format(norm_mes['111']/(norm_mes['111']+norm_mes['011'])))print("Considered {} samples ".format(norm_mes['111']+norm_mes['011']))print ("Acceptance of {:.1f}% of the samples".format((norm_mes['111']+norm_mes['011'])/100))
We repeated our quantum circuit 10,000 times, but only around 200 instances were female children. This reflects that our data that contains only very few female children.
It results in only 2% of the samples being useful. This can become a problem if the network becomes complex.
What if we had a way of increasing the probability of obtaining the proper evidence?
We have one! In a QBN, a quantum state represents a possible configuration of passenger ...