Quantum hardware has been advancing slowly but surely, and year by year we are reaching higher usable qubit counts. In the future, sufficiently large and reliable processors may support more complex machine learning models, including neural networks. To exploit such hardware, algorithms native to quantum devices are required.

One promising direction is quantum convolutional neural networks (QCNNs). Inspired by classical CNNs, QCNNs may offer advantages in cases where input data is inherently quantum, such as states from quantum experiments. However, available qubits remain limited, and requirements strongly depend on encoding strategies.

The goal is to design QCNNs that use quantum resources more efficiently while maintaining predictive power.

Tasks to be performed by the student will include:

Overview and summarize key research on QCNNs, qubit reuse/recycling, quantum machine learning, and resource-efficient neural network designs.
Identify or construct suitable datasets, including synthetic and quantum data.
Develop QCNN architectures with qubit reuse and test different encoding strategies.
Run inference experiments on available quantum devices.
Evaluate performance against standard QCNNs and comparable to classical CNNs.
Provide detailed documentation of the methods and results, summarize contributions, and discuss future research directions.