This website requires JavaScript.

Transformers as Strong Lens Detectors- From Simulation to Surveys

Hareesh ThuruthipillyMargherita GrespanAdam Zadro\.zny
Dec 2022
摘要
With the upcoming large-scale surveys like LSST, we expect to findapproximately $10^5$ strong gravitational lenses among data of many orders ofmagnitude larger. In this scenario, the usage of non-automated techniques istoo time-consuming and hence impractical for science. For this reason, machinelearning techniques started becoming an alternative to previous methods. Wepropose a new machine learning architecture, based on the principle ofself-attention, trained to find strong gravitational lenses on simulated datafrom the Bologna Lens Challenge. Self-attention-based models have clearadvantages compared to simpler CNNs and highly competing performance incomparison to the current state-of-art CNN models. We apply the proposed modelto the Kilo Degree Survey, identifying some new strong lens candidates,however, these have been identified among a plethora of false positives whichmade the application of this model not so advantageous. Therefore, throughoutthis paper, we investigate the pitfalls of this approach, and possiblesolutions, such as transfer learning, are proposed.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答