We have demonstrated a general approach for imposing physically motivated inductive biases on Graph Networks and Hamiltonian GNs to learn interpretable representations, and potentially improved zero-shot generalization

We have generated simulated images of galaxies which have been lensed by dark matter halos with substructure in order to see if machine learning techniques can tell between different kinds of substructure

We provide physical explanations for the dependences, and model the effect of the halo environment on its Hi mass using machine learning tools like random forests and symbolic regression

We investigate the use of approximate Bayesian neural networks in modeling hundreds of time-delay gravitational lenses for Hubble constant determination

As the 21-cm signal parameters are closely associated with the physics of the evolution of the signal, the prediction of the signal parameters gives us a direct physical interpretation of the signal

gamma-ray bursts would act as perfect standard candles if correlations between GRB photometric and spectroscopic properties would somehow be related to GRB intrinsic properties

With the success of machine learning and especially Convolutional Neural Network in image processing, we investigated a new CNN based technique to estimate the photo-z of a galaxy

While the application of machine learning for the prediction of cosmological parameters in the context of weak lensing is not a new concept per se, our work differs from the rest of the literature

For both Machine learning approaches we find that the reconstructed errors are consistent with each other, we are confident in our reconstruction as the Genetic Algorithms and the Gaussian Processes are in principle rather different reconstruction methods

We present ForSE, a novel Python package which aims at overcoming the current limitations in the simulation of diffuse Galactic radiation, in the context of Cosmic Microwave Background experiments

We found 158k quasar candidates with minimum classification probability p(QSOcand) > 0.9 at r < 22, and a total of 311k quasar candidates with p(QSOcand) > 0.98 for r < 23.5, i.e. in the extension to the close extrapolation data

We found that neural networks show promising results on detecting multiple dark matter subhalos, and learn to reject the subhalos on the lensing arc of a smooth lens where there is no subhalo

We present FlowPM - a cosmological N-body code implemented in Mesh-TensorFlow for GPU-accelerated, distributed, and differentiable simulations to tackle the unprecedented modeling and analytic challenges posed by the generation of cosmological surveys

The mock observation and source are shown in fig. 1, along with the mean observation and source reconstructed with a five-layer GP source model in our first analysis step

We have shown how a Generative Adversarial Networks super-resolution architecture inspired by StyleGAN2 can be used to enhance cosmological simulations so that they reproduce the appearance and statistics of much higherresolution models