qiyaoliang/Quantum-Deep-Learning

Recent advances in many fields have accelerated the demand for classification, regression, and detection problems from few 2D images/projections. Often, the heart of these modern techniques utilize neural networks, which can be implemented with deep learning algorithms. In our neural network architecture, we embed a dynamically programmable quantum circuit, acting as a hidden layer, to learn the correct parameters to correctly classify handwritten digits from the MNIST database. By starting small and making incremental improvements, we successfully reach a stunning ~95% accuracy on identifying previously unseen digits from 0 to 7 using this architecture!

Jupyter Notebook