Table 3 Results of the models tested. The rows are grouped by model, with blue indicating the QNN model, green indicating the HQNN model, and red indicating the classical models. The best result for each metric and within each model is underlined. The name Hanoi refers to the IBM “Hanoi” quantum computer, a 27-qubit system. * Ignores the parameters used for generating the embedding vector before the QNN training started. ** Model started with a “head start” of 5 epochs of classically simulated training. *** TF-IDF threshold was found by stepping though the values from 0 to 1 in steps of 0.01 and then checking for the best accuracy score on the training set.

From: Hybrid quantum neural networks show strongly reduced need for free parameters in entity matching