Download PDFOpen PDF in browserQuantum Intelligence : One Neural Network Trains Another Using Its Prior KnowledgeEasyChair Preprint 23749 pages•Date: January 13, 2020AbstractArtificial Neural networks are capable of learning for themselves that will be essential for the future quantum computers mainly quantum error correction and these networks outstrip other error correction strategies. The quantum information is highly sensitive to noise from its environment and needs regular quantum error correction, and this role is performed by artificial neural networks as they gather information about the state of the quantum bits. The solution is in the form of an additional neural network i.e. one neural network uses its prior knowledge of the quantum computer that is to be controlled to train another and guide towards successful quantum correction. In this paper, we propose both conventional neural network model and a novel deep network structure i.e. “ Network In Network” ( NIN) to address quantum error correction. The conventional convolutional layer uses linear filters followed by nonlinear activation function to scan the input. Instead, Deep NIN uses multiple micro neural network to enhance local modelling and utilize global average pooling over feature maps in the classification layer, which is less prone to overfitting than traditional fully connected layers. We demonstrated the performances with MNIST dataset and the test results are encouraging which motivates the possibility of one neural network training another via NIN. Keyphrases: Artificial Neural Networks, Network in Network, Quantum Computers, Quantum intelligence, prior knowledge
|