QuaN: Noisy Dataset For Quantum Machine Learning

QuaN is a collection of specially designed datasets for exploring the impact of noise quantum machine learning and other applications. The presented work focuses on the transformation of clean datasets into noisy counterparts across diverse domains, including MNIST-handwritten digits datasets, Medical MNIST, IRIS datasets and Mobile Health datasets. The dataset is created using noise from classical and quantum domains. The classical noise includes Gaussian distribution, Salt and Pepper method, Random Perturbation, Class Imbalance, and Missing values whereas the quantum noise includes bitflip, phase flip, amplitude damping etc. The dataset is stored in encoded as NumPy array for classical noise and quantum circuit for quantum noise which can be directly loaded and utilized for a QML application. PennyLane is used to create the dataset for tasks such as data encoding and Qasm circuit creation.