Multiclass classification using quantum convolutional neural networks with hybrid quantum-classical learning

Multiclass classification is of great interest for various applications, example, it a common task in computer vision, where one needs to categorize an image into three or more classes. Here we propose quantum machine learning approach based on convolutional neural networks solving the multiclass problem. The corresponding procedure implemented via TensorFlowQuantum as hybrid quantum-classical (variational) model, output results are fed softmax activation function with subsequent minimization cross entropy loss optimizing parameters circuit. Our conceptional improvements here include new model perceptron and optimized structure We use proposed solve 4-class problem case MNIST dataset using eight qubits data encoding four ancilla qubits; previous have been obtained 3-class problems. show that accuracies our solution similar classical comparable numbers trainable parameters. expect finding provide step towards relevant problems NISQ era beyond.