Search for: All records

This paper presents the Learning the Universe Implicit Likelihood Inference (LtU-ILI) pipeline, a codebase for rapid, user-friendly, and cutting-edge machine learning (ML) inference in astrophysics and cosmology. The pipeline includes software for implementing various neural architectures, training schema, priors, and density estimators in a manner easily adaptable to any research workflow. It includes comprehensive validation metrics to assess posterior estimate coverage, enhancing the reliability of inferred results. Additionally, the pipeline is easily parallelizable, designed for efficient exploration of modeling hyperparameters. To demonstrate its capabilities, we present real applications across a range of astrophysics and cosmology problems, such as: estimating galaxy cluster masses from X-ray photometry; inferring cosmology from matter power spectra and halo point clouds; characterising progenitors in gravitational wave signals; capturing physical dust parameters from galaxy colors and luminosities; and establishing properties of semi-analytic models of galaxy formation. We also include exhaustive benchmarking and comparisons of all implemented methods as well as discussions about the challenges and pitfalls of ML inference in astronomical sciences. All code and examples are made publicly available at https://github.com/maho3/ltu-ili. 

Abstract As the next generation of large galaxy surveys come online, it is becoming increasingly important to develop and understand the machine-learning tools that analyze big astronomical data. Neural networks are powerful and capable of probing deep patterns in data, but they must be trained carefully on large and representative data sets. We present a new “hump” of the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project: CAMELS-SAM, encompassing one thousand dark-matter-only simulations of (100h⁻¹cMpc)³with different cosmological parameters (Ω_mandσ₈) and run through the Santa Cruz semi-analytic model for galaxy formation over a broad range of astrophysical parameters. As a proof of concept for the power of this vast suite of simulated galaxies in a large volume and broad parameter space, we probe the power of simple clustering summary statistics to marginalize over astrophysics and constrain cosmology using neural networks. We use the two-point correlation, count-in-cells, and void probability functions, and we probe nonlinear and linear scales across 0.68 <R<27h⁻¹cMpc. We find our neural networks can both marginalize over the uncertainties in astrophysics to constrain cosmology to 3%–8% error across various types of galaxy selections, while simultaneously learning about the SC-SAM astrophysical parameters. This work encompasses vital first steps toward creating algorithms able to marginalize over the uncertainties in our galaxy formation models and measure the underlying cosmology of our Universe. CAMELS-SAM has been publicly released alongside the rest of CAMELS, and it offers great potential to many applications of machine learning in astrophysics:https://camels-sam.readthedocs.io. 

Virtual reality users are susceptible to disorientation, particularly when using locomotion interfaces that lack self-motion cues. Environmental cues, such as boundaries defined by walls or a fence, provide information to help the user remain oriented. This experiment evaluated whether the type of boundary impacts its usefulness for staying oriented. Participants wore a head-mounted display and performed a triangle completion task in virtual reality by traveling two outbound path segments before attempting to point to the path origin. The task was completed with two teleporting interfaces differing in the availability of rotational self-motion cues, and within five virtual environments differing in the availability and type of boundaries. Pointing errors were highest in an open field without environmental cues, and lowest in a classroom with walls and landmarks. Environments with a single square boundary defined by a fence, drop-off, or floor texture discontinuity led to errors in between the open field and the classroom. Performance with the floor texture discontinuity was similar to that with navigational barriers (i.e., fence and drop-off), indicating that an effective barrier need not be a navigational impediment. These results inform spatial cognitive theory about boundary-based navigation and inform application by specifying the types of environmental and self-motion cues that designers of virtual environments should include to reduce disorientation in virtual reality. 

The number of people who own a virtual reality (VR) head-mounted display (HMD) has reached a point where researchers can readily recruit HMD owners to participate remotely using their own equipment. However, HMD owners recruited online may differ from the university community members who typically participate in VR research. HMD owners (n=220) and non-owners (n=282) were recruited through two online work sites-Amazon's Mechanical Turk and Prolific-and an undergraduate participant pool. Participants completed a survey in which they provided demographic information and completed measures of HMD use, video game use, spatial ability, and motion sickness susceptibility. In the context of the populations sampled, the results provide 1) a characterization of HMD owners, 2) a snapshot of the most commonly owned HMDs, 3) a comparison between HMD owners and non-owners, and 4) a comparison among online workers and undergraduates. Significant gender differences were found: men reported lower motion sickness susceptibility and more video game hours than women, and men outperformed women on spatial tasks. Men comprised a greater proportion of HMD owners than non-owners, but after accounting for this imbalance, HMD owners did not differ appreciably from non-owners. Comparing across recruitment platform, male undergraduates outperformed male online workers on spatial tests, and female undergraduates played fewer video game hours than female online workers. The data removal rate was higher from Amazon compared to Prolific, possibly reflecting greater dishonesty. These results provide a description of HMD users that can inform researchers recruiting remote participants through online work sites. These results also signal a need for caution when comparing in-person VR research that primarily enrolls undergraduates to online VR research that enrolls online workers. 

Teleporting is a popular interface for locomotion through virtual environments (VEs). However, teleporting can cause disorientation. Spatial boundaries, such as room walls, are effective cues for reducing disorientation. This experiment explored the characteristics that make a boundary effective. All boundaries tested reduced disorientation, and boundaries representing navigational barriers (e.g., a fence) were no more effective than those defined only by texture changes (e.g., flooring transition). The findings indicate that boundaries need not be navigational barriers to reduce disorientation, giving VE designers greater flexibility in the spatial cues to include. 

Teleporting interfaces are widely used in virtual reality applications to explore large virtual environments. When teleporting, the user indicates the intended location in the virtual environment and is instantly transported, typically without self-motion cues. This project explored the cost of teleporting on the acquisition of survey knowledge (i.e., a ”cognitive map”). Two teleporting interfaces were compared, one with and one without visual and body-based rotational self-motion cues. Both interfaces lacked translational self-motion cues. Participants used one of the two teleporting interfaces to find and study the locations of six objects scattered throughout a large virtual environment. After learning, participants completed two measures of cognitive map fidelity: an object-to-object pointing task and a map drawing task. The results indicate superior spatial learning when rotational self-motion cues were available. Therefore, virtual reality developers should strongly consider the benefits of rotational self-motion cues when creating and choosing locomotion interfaces. 

The AKLT spin chain is the prototypical example of a frustration-free quantum spin system with a spectral gap above its ground state. Affleck, Kennedy, Lieb, and Tasaki also conjectured that the two-dimensional version of their model on the hexagonal lattice exhibits a spectral gap. In this pa- per, we introduce a family of variants of the two-dimensional AKLT model depending on a positive integer n, which is defined by decorating the edges of the hexagonal lattice with one-dimensional AKLT spin chains of length n. We prove that these decorated models are gapped for all n ≥ 3.