Fig. 8
From: Post-variational classical quantum transfer learning for binary classification
Accuracy comparison of CQTL model in qiskit and Pennylane sdk with ResNet50 as the pre-trained model. These are the hyper parameters that was selected for both Pennylane and qiskit performance assessment on CQTL model, batch size: 32, number of qubits: 8, loss function is cross entropy loss function and learning rate: 0.1 with a step size of 10. epochs: 30. CQTL model Implementation in Pennylane was with 3 layers of parameterized ansatz where as in qiskit it was just one layer. Overall time consumed including 3 reps of CQTL in Pennylane is equal to 1 layer of qiskit.
