Fig. 1: Schematic diagram of AdaBoost.Q.
From: Quantum ensemble learning with a programmable superconducting processor
The algorithm is designed to generate a strong quantum classifier (right) by combining multiple weak quantum classifiers (middle). Each weak quantum classifier takes the same training data as an input, and it outputs the classification result of each data sample along with a probability P, which characterizes the confidence of the prediction. The weak quantum classifiers are trained iteratively, using reweighted versions of the training set, with the weights depending on the correctness of the predictions and finely tuned by the output probabilities of the previous classifier. This allows the subsequent quantum classifiers to focus on samples that were not well classified previously. The sample weights for the first classifier are assigned evenly among the training set.
