On the Performance of Various Deep Transfer Learning CNN Models in Glitch Waveform Identification in Gravitational-Wave Data.
Year: 2021
Type: preprint
Abstract: LIGO is considered the most sensitive and complicated gravitational experiment ever built. Its main objective is to detect the gravitational wave from the strongest events in the universe by observing if the length of its 4-kilometer arms change by a distance 10,000 times smaller than the diameter of a proton. Due to its sensitivity, LIGO is prone to the disturbance of external noises which affects the data being collected to detect the gravitational wave. These noises are commonly called by the LIGO community as glitches. The general objective of the study is to evaluate the performance of various deep transfer learning models namely ResNet101, ResNet101V2, ResNet152, ResNet50, ResNet50v2, VGG16, VGG19, Xception, InceptionResnetV2, and DenseNet169 in glitch waveform detection in gravitational-wave data. The model VGG19 recorded the hight AUC-ROC with 98.98%. On the other hand, ResNet152 recorded the lowest AUC-ROC of 93.954% which performs poorly in identifying almost half of the classes in the dataset. It is also observed that less complex model like VGG19,DenseNet169 and VGG16 performs better than most of the more complex models which might indicate that less complex models might be preferred when identifying glitch waveforms.
Source: arXiv (Cornell University)
Authors: Reymond Mesuga, Brian James Bayanay
Language: English
Subfield: Astronomy and Astrophysics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3209235582