| 講演抄録／キーワード |
| 講演名 |
2021-11-30 13:30
［ポスター講演］Machine Learning techniques for unitary design classification:A comparative study
○Yaswitha Gujju・Bo Yang・Dr. Yuko Kuroki・Dr. Hiroshi Imai（UTokyo） |
| 抄録 |
（和） |
The recent use of correlator functions to identify the degree of pseudorandomness in a qubit system opens up immense possibilities to the usage of machine learning in the field of quantum information. In this research we try to deepen our understanding of how the 4-point OTOC function encodes information across different systems containing 3,4,7 and 8 qubits. We employ unsupervised dimensionality reducing technique to visualize the data sets for the different qubit systems and implement a multi-class classification model to discriminate based on the order of the pseudorandom operators for each of the systems individually. Additionally, we examine the possibility of extrapolating the performance of the models trained on a smaller qubit system (3 and 4 qubits) to relatively higher qubit system (7 and 8 qubits) to reduce the need to train the model repeatedly for every qubit system given that the size of the data set varies with the number of qubits. For this, we make use of feature extraction techniques to have a consistent feature dimension across all qubit systems. |
| （英） |
The recent use of correlator functions to identify the degree of pseudorandomness in a qubit system opens up immense possibilities to the usage of machine learning in the field of quantum information. In this research we try to deepen our understanding of how the 4-point OTOC function encodes information across different systems containing 3,4,7 and 8 qubits. We employ unsupervised dimensionality reducing technique to visualize the data sets for the different qubit systems and implement a multi-class classification model to discriminate based on the order of the pseudorandom operators for each of the systems individually. Additionally, we examine the possibility of extrapolating the performance of the models trained on a smaller qubit system (3 and 4 qubits) to relatively higher qubit system (7 and 8 qubits) to reduce the need to train the model repeatedly for every qubit system given that the size of the data set varies with the number of qubits. For this, we make use of feature extraction techniques to have a consistent feature dimension across all qubit systems. |
| キーワード |
（和） |
Haar measure / Quantum Information / Unitary design / Machine Learning / Unsupervised Learning / / / |
| （英） |
Haar measure / Quantum Information / Unitary design / Machine Learning / Unsupervised Learning / / / |
| 文献情報 |
信学技報 |
| 資料番号 |
|
| 発行日 |
|
| ISSN |
|
| PDFダウンロード |
|