The following are modifications or improvements to original notebooks. Please refer to the authors' models for the published primary work. Algorithm HRyERyT1 ResNet50 had the lowest test loss and high test accuracy of 96.7%. HRyERyT6 and Author's Algorithm ResNet18 had higher test loss, but with the high test accuracy. HRyERyT8 and HRyE ResNet18 had higher loss, with an accuracy of 96.1%. Increasing epochs from 25 to 30 did not change any of the best test values. Raising the batch size with no reduction in learning rate yielded the lowest loss of 0.136 with 96.7% accuracy - Therefore Algorithm HRyERyT1 ResNet50 should be utilized first for this 4 qubit model with other binary datasets.

Based on the 'Quantum Transfer Learning' PennyLane model. The author's original model accuracy was 85.0%. All runs were performed with default.qubit and a V100 GPU in a Colab notebook. The 2 Class PyTorch Hymenoptera Dataset downloads using this Link. A classical transfer learning model at 25 epochs achieved 95.4% Accuracy for the 'fine tuning' setting and the 'feature extractor' setting which were faster than quantum-classical runs.