Spring 2023 Physics Colloquium:- Searching gravitational wave signals with autoregressive approach and deep learning
The speaker and the team are experimenting novel approaches of searching gravitational wave signal by employing autoregressive modeling and deep learning. The Autoregressive Integrated Moving Average(ARIMA) model holds the potential for an automatic pipeline for noise reduction, event candidate detections and template-free parameter estimation. In their proof-of-concept experiment with LIGO data, they demonstrated that their proposed scheme can extract the waveform with higher fidelity than the conventional spectral whitening. They have also proposed a deep learning algorithm by incorporating Generative Adversarial Network(GAN). By devising a GAN-based data augmentation scheme, they can tackle both small sample problem and imbalance class problem in classifying gravitational wave spectrograms with deep networks. By experimenting this framework in classifying glitches in the Gravity Spy dataset, they found that their method can yield a performance comparable to or even exceed that achieved by transfer-learning. Such technique can also increase the diversity of training data which can facilitate the search of gravitational wave signals with stochastic nature (e.g. from core-collapsed supernovae).
時間：2023.02.01(三) 14:00~15:00
地點：香港大學 明華綜合大樓 3字樓 MW325
講者：Prof. HUI Chung Yue (韓國忠南國立大學 天文與空間科學院)
語言：英語
免費講座，歡迎參加
【此屬轉載訊息，以主事單位發佈為準】
此活動由香港大學物理系主辦。
