Cosmology with Bayesian Neural Networks

几十年来，暗能量和暗物质的性质一直困扰着宇宙学界。即将到来的像欧几里德这样的第四阶段星系调查将提供有关非线性结构的详细数据。我们将利用这些新信息来约束暗能量和修正的引力理论。我们正在开发一个贝叶斯神经网络，它使用包括非线性模式在内的物质功率谱来推导出最有可能控制底层宇宙的引力理论。稍后，我们将开发一种类似的可观测物质分类技术，可以从欧几里德数据获得。

The nature of Dark Energy and Dark Matter have been puzzling the cosmological community for decades. Upcoming stage IV galaxy surveys like Euclid will provide detailed data on non-linear structures. We will use this new information to constrain dark energy and modified gravity theories.We are developing a Bayesian Neural Network that uses the matter power spectrum including the non-linear modes to deduce the theory of gravity that is most likely to govern the underlying universe.Later on we are going to develop a similar classification technique for observables that can be obtained from Euclid data.