Discovering the building blocks of dark matter halo density profiles with neural networks

Lucie-Smith, Luisa; Peiris, Hiranya V.; Pontzen, Andrew; Nord, B.; Thiyagalingam, Jeyan; Piras, Davide

doi:10.1103/physrevd.105.103533

Cited by 14 publications

(8 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We will also reproduce the results of Lucie-Smith et al [32] in Sect. IV 2, for which the architecture is slightly different: the latent samples are combined with a given query (the radius r) and fed through the decoder to predict dark matter halo density profiles at each given r. This model is referred to as the interpretable variational encoder (IVE), with an analogous loss function to Eq.…”

Section: Representation Learningmentioning

confidence: 60%

“…Dark matter halos forming within such simulations exhibit a universal spherically-averaged density profile as a function of their radius [106][107][108]; this universality encompasses a huge range of halo masses and persists within different cosmological models. While the universality of the density profile is still not fully understood, Lucie-Smith et al [32,LS22 hereafter] showed that it is possible to train a deep representation learning model to compress raw dark matter halo data into , with bandwidths of 0.3 (lower limit) and 0.1 (upper limit). There is good agreement between the two approaches; in particular, the GMM-MI estimates overlap with the KDE estimates at lower (higher) bandwidth when the MI estimates are higher (lower), due to the different KDE bandwidths sometimes underfitting and sometimes overfitting the data.…”

Section: Dark Matter Halo Density Profilesmentioning

confidence: 99%

“…MI has found applications in a variety of scientific fields, including astrophysics [25][26][27][28][29][30][31][32][33], biophysics [34][35][36][37][38][39][40], and dynamical complex systems [41][42][43][44][45][46][47][48], to name a few. However, estimating the mutual information I(X, Y ) between two random variables X and Y , given samples from their joint distribution p (X,Y ) , remains a long-standing challenge, since it requires p (X,Y ) to be known or estimated accurately [49,50].…”

Section: Introductionmentioning

confidence: 99%

“…( 1), are heavily dependent on the training hyperparameters, and have been shown to suffer from a poor variance-bias tradeoff [75]. The use of MI estimates for interpreting deep representation learning has recently been investigated as well [19,32,79,80]; however, exploiting MI to interpret deep representation learning requires the uncertainties on the MI estimate to be quantified, ensuring that any trends in MI are statistically significant.…”

Section: Introductionmentioning

confidence: 99%

“…7. Top panel:Validation log-likelihood per sample as a function of the number of Gaussian mixture model (GMM) components, when fitting joint samples from the ground-truth density for the first radial bin and the second latent variable of the IVE infall model[32]. The GMM reaches the highest log-likelihood at 7 components (dashed black line), which is the value used to then estimate mutual information according to the procedure described in Sect.…”

mentioning

confidence: 99%

See 4 more Smart Citations

A robust estimator of mutual information for deep learning interpretability

Piras,

Peiris,

Pontzen

et al. 2022

Preprint

View full text Add to dashboard Cite

We develop the use of mutual information (MI), a well-established metric in information theory, to interpret the inner workings of deep learning models. To accurately estimate MI from a finite number of samples, we present GMM-MI (pronounced "Jimmie"), an algorithm based on Gaussian mixture models that can be applied to both discrete and continuous settings. GMM-MI is computationally efficient, robust to the choice of hyperparameters and provides the uncertainty on the MI estimate due to the finite sample size. We extensively validate GMM-MI on toy data for which the ground truth MI is known, comparing its performance against established mutual information estimators. We then demonstrate the use of our MI estimator in the context of representation learning, working with synthetic data and physical datasets describing highly non-linear processes. We train deep learning models to encode high-dimensional data within a meaningful compressed (latent) representation, and use GMM-MI to quantify both the level of disentanglement between the latent variables, and their association with relevant physical quantities, thus unlocking the interpretability of the latent representation. We make GMM-MI publicly available in this GitHub repository.

show abstract

Section: Representation Learningmentioning

confidence: 60%

Section: Dark Matter Halo Density Profilesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

A robust estimator of mutual information for deep learning interpretability

Piras,

Peiris,

Pontzen

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

The Dawes Review 10: The impact of deep learning for the analysis of galaxy surveys

Huertas-Company

Lanusse²

2023

Publ. Astron. Soc. Aust.

View full text Add to dashboard Cite

The amount and complexity of data delivered by modern galaxy surveys has been steadily increasing over the past years. New facilities will soon provide imaging and spectra of hundreds of millions of galaxies. Extracting coherent scientific information from these large and multi-modal data sets remains an open issue for the community and data-driven approaches such as deep learning have rapidly emerged as a potentially powerful solution to some long lasting challenges. This enthusiasm is reflected in an unprecedented exponential growth of publications using neural networks, which have gone from a handful of works in 2015 to an average of one paper per week in 2021 in the area of galaxy surveys. Half a decade after the first published work in astronomy mentioning deep learning, and shortly before new big data sets such as Euclid and LSST start becoming available, we believe it is timely to review what has been the real impact of this new technology in the field and its potential to solve key challenges raised by the size and complexity of the new datasets. The purpose of this review is thus two-fold. We first aim at summarising, in a common document, the main applications of deep learning for galaxy surveys that have emerged so far. We then extract the major achievements and lessons learned and highlight key open questions and limitations, which in our opinion, will require particular attention in the coming years. Overall, state-of-the-art deep learning methods are rapidly adopted by the astronomical community, reflecting a democratisation of these methods. This review shows that the majority of works using deep learning up to date are oriented to computer vision tasks (e.g. classification, segmentation). This is also the domain of application where deep learning has brought the most important breakthroughs so far. However, we also report that the applications are becoming more diverse and deep learning is used for estimating galaxy properties, identifying outliers or constraining the cosmological model. Most of these works remain at the exploratory level though which could partially explain the limited impact in terms of citations. Some common challenges will most likely need to be addressed before moving to the next phase of massive deployment of deep learning in the processing of future surveys; for example, uncertainty quantification, interpretability, data labelling and domain shift issues from training with simulations, which constitutes a common practice in astronomy.

show abstract

From inflation to dark matter halo profiles: the impact of primordial non-Gaussianities on the central density cusp

Stahl,

Mai,

Famaey

et al. 2024

J. Cosmol. Astropart. Phys.

View full text Add to dashboard Cite

It has recently been shown that local primordial non-Gaussianities (PNG) with significant amplitude (|f NL| ∼ 1000), at small (Mpc) scales, can help in forming simulated galaxies with more disky baryonic kinematics than in the Gaussian case, while generating matter power spectra that can differ by up to 20% from the Gaussian case at non-linear scales. Here, we explore in detail the consequences of such small-scale PNG on the dark matter halo profiles. We show in particular that, for negative f NL, dark matter halos formed in collisionless simulations are not always well described by the traditional Navarro-Frenk-White (NFW) profiles, as supported by their sparsity distribution. We conclude that NFW profiles are not as clear attractors for the density profiles of dark matter halos in the presence of PNG than in the case of a Gaussian contrast density field. We show how alternatives to the NFW profile can describe halos both in the Gaussian and non-Gaussian cases. From the combination of our sparsity analysis and the quality of the adjustments of the density profiles with a minimal extension to NFW, we conclude that z = 1 halos carry the most interesting information about PNG.

show abstract

Discovering the building blocks of dark matter halo density profiles with neural networks

Cited by 14 publications

References 37 publications

A robust estimator of mutual information for deep learning interpretability

A robust estimator of mutual information for deep learning interpretability

The Dawes Review 10: The impact of deep learning for the analysis of galaxy surveys

From inflation to dark matter halo profiles: the impact of primordial non-Gaussianities on the central density cusp

Contact Info

Product

Resources

About