We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Quantity study on a novel quantum neural network with alternately controlled gates for binary image classification.
- Authors
Bai, Qi; Hu, Xianliang
- Abstract
A novel quantum neural network(QNN) is proposed, in which quantum probability image encoding(QPIE) and specially designed ansatz are used. QPIE can exponentially reduce qubits for image encoding by using quantum superposition. The parameter gates in ansatz are selected from the universal gate set for quantum computing, which guarantees the expressibility of models. The proposed QNN can be trained by supervised learning. In this article, various experiments are conducted to explore the factors that affect accuracy. The results derive from MNIST show that both the improvement of resolution and the repetition of layers have a positive contribution to accuracy. The enhancement of the expressibility of a single layer by replacing CX gates with R y gates also improves the performance of the model.
- Subjects
IMAGE recognition (Computer vision); SUPERVISED learning; QUANTUM superposition; QUANTUM computing; QUANTUM gates; QUBITS; QUANTUM computers
- Publication
Quantum Information Processing, 2023, Vol 22, Issue 5, p1
- ISSN
1570-0755
- Publication type
Article
- DOI
10.1007/s11128-023-03929-y