@techreport{oai:ipsj.ixsq.nii.ac.jp:00240380,
 author = {三輪, 拓真 and 小田, 悠介 and 河野, 誠也 and 吉野, 幸一郎},
 issue = {15},
 month = {Oct},
 note = {様々な応用可能性と量子計算への相性の良さから，量子畳み込みニューラルネットワーク (QCNN) への注目が高まっている．量子畳み込み学習は性能面で優れているものの，重みフィルタの数だけ量子回路を呼び出す必要があり，計算効率に課題があった．本研究では，この重みフィルタ数とチャネル数の関係に着目し，計算アルゴリズムを効率化することを提案する．具体的には，複数のパウリ演算を観測して効率的に量子状態を次の層に残すことで，チャネル数の削減を行った．この手法により，精度を維持しつつパラメータ数及び計算時間を大きく削減できることを確認した．, Due to its various potential applications and suitability for quantum computing, there is growing interest in quantum convolutional neural networks (QCNNs). Although QCNN is superior in terms of performance, it has the problem of computational efficiency, as it requires many calls of a quantum circuit corresponding in proportion to the number weight filters. In this study, we propose an efficient computational algorithm by focusing on the relationship between the number of weight filters and the number of channels. Specifically, we reduced the number of channels by efficiently remain the quantum state in the next layer by observing multiple Pauli operations. We confirmed that this method can greatly reduce the number of parameters and calculation time while maintaining accuracy.},
 title = {QCNNにおける効率的なチャネル計算アルゴリズム},
 year = {2024}
}