CosmoGAN: creating high-fidelity weak lensing convergence maps using Generative Adversarial Networks

Mustafa, Mustafa; Bard, Deborah; Bhimji, W.; Lukić, Zarija; Al‐Rfou, Rami; Kratochvíl, Jan

doi:10.1186/s40668-019-0029-9

Cited by 134 publications

(105 citation statements)

References 40 publications

(48 reference statements)

Supporting

Mentioning

102

Contrasting

Order By: Relevance

“…GANs were proposed for generating of matter distributions in 2D. A generative model for the projected matter distribution, also called a mass map, was introduced by [7]. Mass maps are cosmological observables, as they are reconstructed by techniques such as, for example, gravitational lensing [20].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Cosmological N-body simulations: a challenge for scalable generative models

Perraudin

Srivastava

Lucchi

et al. 2019

Comput. Astrophys. Cosmol.

View full text Add to dashboard Cite

Deep generative models, such as Generative Adversarial Networks (GANs) or Variational Autoencoders (VAs) have been demonstrated to produce images of high visual quality. However, the existing hardware on which these models are trained severely limits the size of the images that can be generated. The rapid growth of high dimensional data in many fields of science therefore poses a significant challenge for generative models. In cosmology, the large-scale, three-dimensional matter distribution, modeled with N-body simulations, plays a crucial role in understanding of evolution of structures in the universe. As these simulations are computationally very expensive, GANs have recently generated interest as a possible method to emulate these datasets, but they have been, so far, mostly limited to two dimensional data. In this work, we introduce a new benchmark for the generation of three dimensional N-body simulations, in order to stimulate new ideas in the machine learning community and move closer to the practical use of generative models in cosmology. As a first benchmark result, we propose a scalable GAN approach for training a generator of N-body three-dimensional cubes. Our technique relies on two key building blocks, (i) splitting the generation of the high-dimensional data into smaller parts, and (ii) using a multi-scale approach that efficiently captures global image features that might otherwise be lost in the splitting process. We evaluate the performance of our model for the generation of N-body samples using various statistical measures commonly used in cosmology. Our results show that the proposed model produces samples of high visual quality, although the statistical analysis reveals that capturing rare features in the data poses significant problems for the generative models. We make the data, quality evaluation routines, and the proposed GAN architecture publicly available at https://github.com/nperraud/3DcosmoGAN.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Recently, GANs have been proposed for emulating the matter distributions in two dimensions [6,7]. These approaches have been successful in generating data of high visual quality, and almost indistinguishable from the real simulations to experts.…”

Section: Introductionmentioning

confidence: 99%

Cosmological N-body simulations: a challenge for scalable generative models

Perraudin

Srivastava

Lucchi

et al. 2019

Comput. Astrophys. Cosmol.

View full text Add to dashboard Cite

show abstract

“…This network implements a deep convolutional generative adversarial network (DC‐GAN) for producing cosmology mass maps . These are usually obtained in expensive simulations and thus having a network, which can generate cheap surrogate examples, is mandatory for training inference workflows on cosmological parameter estimations.…”

Section: Modelsmentioning

confidence: 99%

“…This network implements a deep convolutional generative adversarial network (DC-GAN) for producing cosmology mass maps. 23 After each (de-)convolutional layer of generator and discriminator, a batch norm layer is inserted to improve precision. Since the network consists of sparse layers and only very lightweight dense layers, it is inherently scalable in a performant way.…”

Section: Cosmoganmentioning

confidence: 99%

TensorFlow at Scale: Performance and productivity analysis of distributed training with Horovod, MLSL, and Cray PE ML

Kurth

Smorkalov²,

Mendygral³

et al. 2018

Concurrency and Computation

View full text Add to dashboard Cite

Summary Deep learning has proven to be a successful tool for solving a large variety of problems in various scientific fields and beyond. In recent years, the models as well as the available datasets have grown bigger and more complicated, and thus, an increasing amount of computing resources is required in order to train these models in a reasonable amount of time. Besides being able to use HPC resources, deep learning model developers want flexible frameworks which allow for rapid prototyping. One of the most important of these frameworks is Google TensorFlow, which provides both features, ie, good performance as well as flexibility. In this paper, we discuss different solutions for scaling the TensorFlow Framework to thousands of nodes on contemporary Cray XC supercomputing systems.

show abstract

“…ML methods can paint galaxies themselves, knowing only information about host dark matter halo (Agarwal et al 2018;Zhang et al 2019). Finally, it is possible to predict directly the evolution of cosmological structure formation (in terms of Press-Schechter theory) (Lucie-Smith et al 2018;He et al 2019) or simulate the Cosmic Web (Rodriguez et al 2018) or weak lensing map (Mustafa et al 2019) via ML .…”

Section: Introductionmentioning

confidence: 99%

Large-scale structures in the ΛCDM Universe: network analysis and machine learning

Tsizh

Novosyadlyj

Holovatch

et al. 2020

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

We perform an analysis of the Cosmic Web as a complex network, which is built on a ΛCDM cosmological simulation. For each of nodes, which are in this case dark matter halos formed in the simulation, we compute 10 network metrics, which characterize the role and position of a node in the network. The relation of these metrics to topological affiliation of the halo, i.e. to the type of large scale structure, which it belongs to, is then investigated. In particular, the correlation coefficients between network metrics and topology classes are computed. We have applied different machine learning methods to test the predictive power of obtained network metrics and to check if one could use network analysis as a tool for establishing topology of the large scale structure of the Universe. Results of such predictions, combined in the confusion matrix, show that it is not possible to give a good prediction of the topology of Cosmic Web (score is ≈ 70 % in average) based only on coordinates and velocities of nodes (halos), yet network metrics can give a hint about the topological landscape of matter distribution.

show abstract

CosmoGAN: creating high-fidelity weak lensing convergence maps using Generative Adversarial Networks

Cited by 134 publications

References 40 publications

Cosmological N-body simulations: a challenge for scalable generative models

Cosmological N-body simulations: a challenge for scalable generative models

TensorFlow at Scale: Performance and productivity analysis of distributed training with Horovod, MLSL, and Cray PE ML

Large-scale structures in the ΛCDM Universe: network analysis and machine learning

Contact Info

Product

Resources

About