Xuan Qi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xuan Qi has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Xue Mei
College of Electrical Engineering and Control Science, Nanjing Tech University, Nanjing, China
4 shared publications - Ryad Chellali
Nanjing Tech University
3 shared publications - Shipin Yang
College of Electrical Engineering and Control Science, Nanjing Tech University, Nanjing, China
2 shared publications - Yi Wei
Nanjing Tech University
2 shared publications - Hao Fu
College of Electrical Engineering and Control Science, Nanjing Tech University, Nanjing 211816, China
1 shared publication - Peng Liu
Key Laboratory of Measurement and Control of CSE, Ministry of Education, School of Automation, Southeast University, Nanjing 210096, P. R. China
1 shared publication
Latest Publications
- 1 September 2024
- Published by IOP Publishing in Research in Astronomy and Astrophysics
- Vol. 24 (9) , 095023
SFNet: Stellar Feature Network with CWT for Stellar Spectra Recognition
Stellar spectral classification is crucial in astronomical data analysis. However, existing studies are often limited by the uneven distribution of stellar samples, posing challenges in practical applications. Even when balancing stellar categories and their numbers, there is room for improvement in classification accuracy. This study introduces a Continuous Wavelet Transform using the Super Morlet wavelet to convert stellar spectra into wavelet images. A novel neural network, the Stellar Feature Network, is proposed for classifying these images. Stellar spectra from Large Sky Area Multi-Object Fiber Spectroscopic Telescope DR9, encompassing five equal categories (B, A, F, G, K), were used. Comparative experiments validate the effectiveness of the proposed methods and network, achieving significant improvements in classification accuracy.
- 1 September 2024
- Published by IOP Publishing in Research in Astronomy and Astrophysics
- Vol. 24 (9) , 095023
SFNet: Stellar Feature Network with CWT for Stellar Spectra Recognition
Stellar spectral classification is crucial in astronomical data analysis. However, existing studies are often limited by the uneven distribution of stellar samples, posing challenges in practical applications. Even when balancing stellar categories and their numbers, there is room for improvement in classification accuracy. This study introduces a Continuous Wavelet Transform using the Super Morlet wavelet to convert stellar spectra into wavelet images. A novel neural network, the Stellar Feature Network, is proposed for classifying these images. Stellar spectra from Large Sky Area Multi-Object Fiber Spectroscopic Telescope DR9, encompassing five equal categories (B, A, F, G, K), were used. Comparative experiments validate the effectiveness of the proposed methods and network, achieving significant improvements in classification accuracy.
- 26 November 2023
- Published by Springer Nature
- p. 528-539
Comparative Analysis of the Linear Regions in ReLU and LeakyReLU Networks
- 26 November 2023
- Published by Springer Nature
- p. 528-539
Comparative Analysis of the Linear Regions in ReLU and LeakyReLU Networks
- 22 September 2023
- Published by Springer Nature
- p. 174-185
Empirical Study on the Effect of Residual Networks on the Expressiveness of Linear Regions
- 22 September 2023
- Published by Springer Nature
- p. 174-185
Empirical Study on the Effect of Residual Networks on the Expressiveness of Linear Regions
- Open Access
- 12 April 2023
- Published by Wiley in Mobile Information Systems
- Vol. 2023, 1-11
A Lightweight Binarized Convolutional Neural Network Model for Small Memory and Low-Cost Mobile Devices
In recent years, the high cost of implementing deep neural networks due to their large model size and parameter complexity has made it a challenging problem to design lightweight models that reduce application costs. The existing binarized neural networks suffer from both the large memory occupancy and the big number of trainable params they use. We propose a lightweight binarized convolutional neural network (CBCNN) model to address the multiclass classification/identification problem. We use both binary weights and activation. We show experimentally that a model using only 0.59 M trainable params is sufficient to reach about 92.94% accuracy on the GTSRB dataset, and it has similar performances compared to other methods on MNIST and Fashion-MNIST datasets. Accordingly, most arithmetic operations with bitwise operations are simplified, thus both used memory size and memory accesses are reduced by 32 times. Moreover, the color information was removed, which also reduced drastically the training time. All these together allow our architecture to run effectively and in real time on simple CPUs (rather than GPUs). Through the results we obtained, we show that despite simplifications and color information removal, our network achieves similar performances compared to classical CNNs with lower costs in both in training and embedded deployment.
- 12 April 2023
- Published by Wiley in Mobile Information Systems
- Vol. 2023, 1-11
Open AccessA Lightweight Binarized Convolutional Neural Network Model for Small Memory and Low-Cost Mobile Devices
In recent years, the high cost of implementing deep neural networks due to their large model size and parameter complexity has made it a challenging problem to design lightweight models that reduce application costs. The existing binarized neural networks suffer from both the large memory occupancy and the big number of trainable params they use. We propose a lightweight binarized convolutional neural network (CBCNN) model to address the multiclass classification/identification problem. We use both binary weights and activation. We show experimentally that a model using only 0.59 M trainable params is sufficient to reach about 92.94% accuracy on the GTSRB dataset, and it has similar performances compared to other methods on MNIST and Fashion-MNIST datasets. Accordingly, most arithmetic operations with bitwise operations are simplified, thus both used memory size and memory accesses are reduced by 32 times. Moreover, the color information was removed, which also reduced drastically the training time. All these together allow our architecture to run effectively and in real time on simple CPUs (rather than GPUs). Through the results we obtained, we show that despite simplifications and color information removal, our network achieves similar performances compared to classical CNNs with lower costs in both in training and embedded deployment.