arXiv subjects

Vanessa Böhm

Publications and source records attributed to Vanessa Böhm.

14 recordsLinked to original sources

arXiv2023

Fast and efficient identification of anomalous galaxy spectra with neural density estimation

Current large-scale astrophysical experiments produce unprecedented amounts of rich and diverse data. This creates a growing need for fast and flexible automated data inspection methods. Deep learning algorithms can capture and pick up subtle variations in rich data sets and are fast to apply once trained. Here, we study the applicability of an unsupervised and probabilistic deep learning framework, the Probabilistic Autoencoder (PAE), to the detection of peculiar objects in galaxy spectra from the SDSS survey. Different to supervised algorithms, this algorithm is not trained to detect a specific feature or type of anomaly, instead it learns the complex and diverse distribution of galaxy spectra from training data and identifies outliers with respect to the learned distribution. We find that the algorithm assigns consistently lower probabilities (higher anomaly score) to spectra that exhibit unusual features. For example, the majority of outliers among quiescent galaxies are E+A galaxies, whose spectra combine features from old and young stellar population. Other identified outliers include LINERs, supernovae and overlapping objects. Conditional modeling further allows us to incorporate additional information. Namely, we evaluate the probability of an object being anomalous given a certain spectral class, but other information such as metrics of data quality or estimated redshift could be incorporated as well. We make our code publicly available at https://github.com/VMBoehm/Spectra_PAE

astro-ph.IM↗

arXiv2022

Reconstructing and Classifying SDSS DR16 Galaxy Spectra with Machine-Learning and Dimensionality Reduction Algorithms

Optical spectra of galaxies and quasars from large cosmological surveys are used to measure redshifts and infer distances. They are also rich with information on the intrinsic properties of these astronomical objects. However, their physical interpretation can be challenging due to the substantial number of degrees of freedom, various sources of noise, and degeneracies between physical parameters that cause similar spectral characteristics. To gain deeper insights into these degeneracies, we apply two unsupervised machine learning frameworks to a sample from the Sloan Digital Sky Survey data release 16 (SDSS DR16). The first framework is a Probabilistic Auto-Encoder (PAE), a two-stage deep learning framework consisting of a data compression stage from 1000 elements to 10 parameters and a density estimation stage. The second framework is a Uniform Manifold Approximation and Projection (UMAP), which we apply to both the uncompressed and compressed data. Exploring across regions on the compressed data UMAP, we construct sequences of stacked spectra which show a gradual transition from star-forming galaxies with narrow emission lines and blue spectra to passive galaxies with absorption lines and red spectra. Focusing on galaxies with broad emission lines produced by quasars, we find a sequence with varying levels of obscuration caused by cosmic dust. The experiments we present here inform future applications of neural networks and dimensionality reduction algorithms for large astronomical spectroscopic surveys.

astro-ph.GA↗

arXiv2022

SAR-based landslide classification pretraining leads to better segmentation

Rapid assessment after a natural disaster is key for prioritizing emergency resources. In the case of landslides, rapid assessment involves determining the extent of the area affected and measuring the size and location of individual landslides. Synthetic Aperture Radar (SAR) is an active remote sensing technique that is unaffected by weather conditions. Deep Learning algorithms can be applied to SAR data, but training them requires large labeled datasets. In the case of landslides, these datasets are laborious to produce for segmentation, and often they are not available for the specific region in which the event occurred. Here, we study how deep learning algorithms for landslide segmentation on SAR products can benefit from pretraining on a simpler task and from data from different regions. The method we explore consists of two training stages. First, we learn the task of identifying whether a SAR image contains any landslides or not. Then, we learn to segment in a sparsely labeled scenario where half of the data do not contain landslides. We test whether the inclusion of feature embeddings derived from stage-1 helps with landslide detection in stage-2. We find that it leads to minor improvements in the Area Under the Precision-Recall Curve, but also to a significantly lower false positive rate in areas without landslides and an improved estimate of the average number of landslide pixels in a chip. A more accurate pixel count allows to identify the most affected areas with higher confidence. This could be valuable in rapid response scenarios where prioritization of resources at a global scale is important. We make our code publicly available at https://github.com/VMBoehm/SAR-landslide-detection-pretraining.

cs.CV↗

arXiv2022

Probabilistic Autoencoder

Principal Component Analysis (PCA) minimizes the reconstruction error given a class of linear models of fixed component dimensionality. Probabilistic PCA adds a probabilistic structure by learning the probability distribution of the PCA latent space weights, thus creating a generative model. Autoencoders (AE) minimize the reconstruction error in a class of nonlinear models of fixed latent space dimensionality and outperform PCA at fixed dimensionality. Here, we introduce the Probabilistic Autoencoder (PAE) that learns the probability distribution of the AE latent space weights using a normalizing flow (NF). The PAE is fast and easy to train and achieves small reconstruction errors, high sample quality, and good performance in downstream tasks. We compare the PAE to Variational AE (VAE), showing that the PAE trains faster, reaches a lower reconstruction error, and produces good sample quality without requiring special tuning parameters or training procedures. We further demonstrate that the PAE is a powerful model for performing the downstream tasks of probabilistic image reconstruction in the context of Bayesian inference of inverse problems for inpainting and denoising applications. Finally, we identify latent space density from NF as a promising outlier detection metric.

cs.LG↗

arXiv2022

Impact of COVID-19 on Astronomy: Two Years In

We study the impact of the COVID-19 pandemic on astronomy using public records of astronomical publications. We show that COVID-19 has had both positive and negative impacts on research in astronomy. We find that the overall output of the field, measured by the yearly paper count, has increased. This is mainly driven by boosted individual productivity seen across most countries, possibly the result of cultural and technological changes in the scientific community during COVID. However, a decreasing number of incoming new researchers is seen in most of the countries we studied, indicating larger barriers for new researchers to enter the field or for junior researchers to complete their first project during COVID. Unfortunately, the overall improvement in productivity seen in the field is not equally shared by female astronomers. By fraction, fewer papers are written by women and fewer women are among incoming new researchers in most countries. Even though female astronomers also became more productive during COVID, the level of improvement is smaller than for men. Pre-COVID, female astronomers in the Netherlands, Australia, Switzerland were equally as or even more productive than their male colleagues. During COVID, no single country's female astronomers were able to be more productive than their male colleagues on average.

astro-ph.IM↗

arXiv2020

MADLens, a python package for fast and differentiable non-Gaussian lensing simulations

We present MADLens a python package for producing non-Gaussian lensing convergence maps at arbitrary source redshifts with unprecedented precision. MADLens is designed to achieve high accuracy while keeping computational costs as low as possible. A MADLens simulation with only $256^3$ particles produces convergence maps whose power agree with theoretical lensing power spectra up to $L{=}10000$ within the accuracy limits of HaloFit. This is made possible by a combination of a highly parallelizable particle-mesh algorithm, a sub-evolution scheme in the lensing projection, and a machine-learning inspired sharpening step. Further, MADLens is fully differentiable with respect to the initial conditions of the underlying particle-mesh simulations and a number of cosmological parameters. These properties allow MADLens to be used as a forward model in Bayesian inference algorithms that require optimization or derivative-aided sampling. Another use case for MADLens is the production of large, high resolution simulation sets as they are required for training novel deep-learning-based lensing analysis tools. We make the MADLens package publicly available under a Creative Commons License (https://github.com/VMBoehm/MADLens).

astro-ph.CO↗

arXiv2019

Lensing corrections on galaxy-lensing cross correlations and galaxy-galaxy auto correlations

We study the impact of lensing corrections on modeling cross correlations between CMB lensing and galaxies, cosmic shear and galaxies, and galaxies in different redshift bins. Estimating the importance of these corrections becomes necessary in the light of anticipated high-accuracy measurements of these observables. While higher order lensing corrections (sometimes also referred to as post Born corrections) have been shown to be negligibly small for lensing auto correlations, they have not been studied for cross correlations. We evaluate the contributing four-point functions without making use of the Limber approximation and compute line-of-sight integrals with the numerically stable and fast FFTlog formalism. We find that the relative size of lensing corrections depends on the respective redshift distributions of the lensing sources and galaxies, but that they are generally small for high signal-to-noise correlations. We point out that a full assessment and judgement of the importance of these corrections requires the inclusion of lensing Jacobian terms on the galaxy side. We identify these additional correction terms, but do not evaluate them due to their large number. We argue that they could be potentially important and suggest that their size should be measured in the future with ray-traced simulations. We make our code publicly available.

astro-ph.CO↗

arXiv2019

Uncertainty Quantification with Generative Models

We develop a generative model-based approach to Bayesian inverse problems, such as image reconstruction from noisy and incomplete images. Our framework addresses two common challenges of Bayesian reconstructions: 1) It makes use of complex, data-driven priors that comprise all available information about the uncorrupted data distribution. 2) It enables computationally tractable uncertainty quantification in the form of posterior analysis in latent and data space. The method is very efficient in that the generative model only has to be trained once on an uncorrupted data set, after that, the procedure can be used for arbitrary corruption types.

stat.ML↗

arXiv2018

Constraining Neutrino Mass with the Tomographic Weak Lensing Bispectrum

We explore the effect of massive neutrinos on the weak lensing shear bispectrum using the Cosmological Massive Neutrino Simulations. We find that the primary effect of massive neutrinos is to suppress the amplitude of the bispectrum with limited effect on the bispectrum shape. The suppression of the bispectrum amplitude is a factor of two greater than the suppression of the small scale power-spectrum. For an LSST-like weak lensing survey that observes half of the sky with five tomographic redshift bins, we explore the constraining power of the bispectrum on three cosmological parameters: the sum of the neutrino mass $\sum m_ν$, the matter density $Ω_m$ and the amplitude of primordial fluctuations $A_s$. Bispectrum measurements alone provide similar constraints to the power spectrum measurements and combining the two probes leads to significant improvements than using the latter alone. We find that the joint constraints tighten the power spectrum $95\%$ constraints by $\sim 32\%$ for $\sum m_ν$, $13\%$ for $Ω_m$ and $57\%$ for $A_s$ .

astro-ph.CO↗

arXiv2018

On the effect of non-Gaussian lensing deflections on CMB lensing measurements

We investigate the impact of non-Gaussian lensing deflections on measurements of the CMB lensing power spectrum. We find that the false assumption of their Gaussianity significantly biases these measurements in current and future experiments at the percent level. The bias is detected by comparing CMB lensing reconstructions from simulated CMB data lensed with Gaussian deflection fields to reconstructions from simulations lensed with fully non-Gaussian deflection fields. The non-Gaussian deflections are produced by ray-tracing through snapshots of an N-body simulation and capture both the non-Gaussianity induced by non-linear structure formation and by multiple correlated deflections. We find that the amplitude of the measured bias is in agreement with analytical predictions by Böhm et al. 2016. The bias is largest in temperature-based measurements and we do not find evidence for it in measurements from a combination of polarization fields ($EB,EB$). We argue that the non-Gaussian bias should be even more important for measurements of cross-correlations of CMB lensing with low-redshift tracers of large-scale structure.

astro-ph.CO↗

arXiv2017

Bayesian weak lensing tomography: Reconstructing the 3D large-scale distribution of matter with a lognormal prior

We present a Bayesian reconstruction algorithm that infers the three-dimensional large-scale matter distribution from the weak gravitational lensing effects measured in the image shapes of galaxies. The algorithm is designed to also work with non-Gaussian posterior distributions which arise, for example, from a non-Gaussian prior distribution. In this work, we use a lognormal prior and compare the reconstruction results to a Gaussian prior in a suite of increasingly realistic tests on mock data. We find that in cases of high noise levels (i.e. for low source galaxy densities and/or high shape measurement uncertainties), both normal and lognormal priors lead to reconstructions of comparable quality, but with the lognormal reconstruction being prone to mass-sheet degeneracy. In the low-noise regime and on small scales, the lognormal model produces better reconstructions than the normal model: The lognormal model 1) enforces non-negative densities, while negative densities are present when a normal prior is employed, 2) better traces the extremal values and the skewness of the true underlying distribution, and 3) yields a higher pixel-wise correlation between the reconstruction and the true density.

astro-ph.CO↗

arXiv2016

Cosmic expansion history from SNe Ia data via information field theory -- the charm code

We present charm (cosmic history agnostic reconstruction method), a novel inference algorithm that reconstructs the cosmic expansion history as encoded in the Hubble parameter $H(z)$ from SNe Ia data. The novelty of the approach lies in the usage of information field theory, a statistical field theory that is very well suited for the construction of optimal signal recovery algorithms. The charm algorithm infers non-parametrically $s(a)=\ln(ρ(a)/ρ_{\mathrm{crit}0})$, the density evolution which determines $H(z)$, without assuming an analytical form of $ρ(a)$ but only its smoothness with the scale factor $a=(1+z)^{-1}$. The inference problem of recovering the signal $s(a)$ from the data is formulated in a fully Bayesian way. In detail, we have rewritten the signal as the sum of a background cosmology and a perturbation. This allows us to determine the maximum a posteriory estimate of the signal by an iterative Wiener filter method. Applying charm to the Union2.1 supernova compilation, we have recovered a cosmic expansion history that is fully compatible with the standard $Λ$CDM cosmological expansion history with parameter values consistent with the results of the Planck mission.

astro-ph.CO↗

arXiv2016

CMB Lensing Beyond the Power Spectrum: Cosmological Constraints from the One-Point PDF and Peak Counts

Unprecedentedly precise cosmic microwave background (CMB) data are expected from ongoing and near-future CMB Stage-III and IV surveys, which will yield reconstructed CMB lensing maps with effective resolution approaching several arcminutes. The small-scale CMB lensing fluctuations receive non-negligible contributions from nonlinear structure in the late-time density field. These fluctuations are not fully characterized by traditional two-point statistics, such as the power spectrum. Here, we use $N$-body ray-tracing simulations of CMB lensing maps to examine two higher-order statistics: the lensing convergence one-point probability distribution function (PDF) and peak counts. We show that these statistics contain significant information not captured by the two-point function, and provide specific forecasts for the ongoing Stage-III Advanced Atacama Cosmology Telescope (AdvACT) experiment. Considering only the temperature-based reconstruction estimator, we forecast 9$σ$ (PDF) and 6$σ$ (peaks) detections of these statistics with AdvACT. Our simulation pipeline fully accounts for the non-Gaussianity of the lensing reconstruction noise, which is significant and cannot be neglected. Combining the power spectrum, PDF, and peak counts for AdvACT will tighten cosmological constraints in the $Ω_m$-$σ_8$ plane by $\approx 30\%$, compared to using the power spectrum alone.

astro-ph.CO↗

arXiv2016

A bias to CMB lensing measurements from the bispectrum of large-scale structure

The rapidly improving precision of measurements of gravitational lensing of the Cosmic Microwave Background (CMB) also requires a corresponding increase in the precision of theoretical modeling. A commonly made approximation is to model the CMB deflection angle or lensing potential as a Gaussian random field. In this paper, however, we analytically quantify the influence of the non-Gaussianity of large-scale structure lenses, arising from nonlinear structure formation, on CMB lensing measurements. In particular, evaluating the impact of the non-zero bispectrum of large-scale structure on the relevant CMB four-point correlation functions, we find that there is a bias to estimates of the CMB lensing power spectrum. For temperature-based lensing reconstruction with CMB Stage-III and Stage-IV experiments, we find that this lensing power spectrum bias is negative and is of order one percent of the signal. This corresponds to a shift of multiple standard deviations for these upcoming experiments. We caution, however, that our numerical calculation only evaluates two of the largest bias terms and thus only provides an approximate estimate of the full bias. We conclude that further investigation into lensing biases from nonlinear structure formation is required and that these biases should be accounted for in future lensing analyses.

astro-ph.CO↗