Showing 1–23 of 23 results for author: Leike, R

The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions

Authors: Eric Wallace, Kai Xiao, Reimar Leike, Lilian Weng, Johannes Heidecke, Alex Beutel

Abstract: Today's LLMs are susceptible to prompt injections, jailbreaks, and other attacks that allow adversaries to overwrite a model's original instructions with their own malicious prompts. In this work, we argue that one of the primary vulnerabilities underlying these attacks is that LLMs often consider system prompts (e.g., text from an application developer) to be the same priority as text from untrus… ▽ More Today's LLMs are susceptible to prompt injections, jailbreaks, and other attacks that allow adversaries to overwrite a model's original instructions with their own malicious prompts. In this work, we argue that one of the primary vulnerabilities underlying these attacks is that LLMs often consider system prompts (e.g., text from an application developer) to be the same priority as text from untrusted users and third parties. To address this, we propose an instruction hierarchy that explicitly defines how models should behave when instructions of different priorities conflict. We then propose a data generation method to demonstrate this hierarchical instruction following behavior, which teaches LLMs to selectively ignore lower-privileged instructions. We apply this method to GPT-3.5, showing that it drastically increases robustness -- even for attack types not seen during training -- while imposing minimal degradations on standard capabilities. △ Less

Submitted 19 April, 2024; originally announced April 2024.

Re-Envisioning Numerical Information Field Theory (NIFTy.re): A Library for Gaussian Processes and Variational Inference

Authors: Gordian Edenhofer, Philipp Frank, Jakob Roth, Reimar H. Leike, Massin Guerdi, Lukas I. Scheel-Platz, Matteo Guardiani, Vincent Eberle, Margret Westerkamp, Torsten A. Enßlin

Abstract: Imaging is the process of transforming noisy, incomplete data into a space that humans can interpret. NIFTy is a Bayesian framework for imaging and has already successfully been applied to many fields in astrophysics. Previous design decisions held the performance and the development of methods in NIFTy back. We present a rewrite of NIFTy, coined NIFTy.re, which reworks the modeling principle, ext… ▽ More Imaging is the process of transforming noisy, incomplete data into a space that humans can interpret. NIFTy is a Bayesian framework for imaging and has already successfully been applied to many fields in astrophysics. Previous design decisions held the performance and the development of methods in NIFTy back. We present a rewrite of NIFTy, coined NIFTy.re, which reworks the modeling principle, extends the inference strategies, and outsources much of the heavy lifting to JAX. The rewrite dramatically accelerates models written in NIFTy, lays the foundation for new types of inference machineries, improves maintainability, and enables interoperability between NIFTy and the JAX machine learning ecosystem. △ Less

Submitted 15 June, 2024; v1 submitted 26 February, 2024; originally announced February 2024.

Comments: 10 pages, 2 figures, published in Journal of Open Source Software (JOSS)

Journal ref: Journal of Open Source Software, volume 9(98), year 2024, page 6593

Measurement in a Unitary World

Authors: Vishal Johnson, Reimar Leike, Philipp Frank, Torsten Enßlin

Abstract: This article explores how measurement can be understood in the context of a universe evolving according to unitary (reversible) quantum dynamics. A unitary measurement procedure is developed consistent with the non-measurement axioms of quantum mechanics wherein the system being measured and the observer become correlated. It is argued that for this to work the correlation necessarily has to be tr… ▽ More This article explores how measurement can be understood in the context of a universe evolving according to unitary (reversible) quantum dynamics. A unitary measurement procedure is developed consistent with the non-measurement axioms of quantum mechanics wherein the system being measured and the observer become correlated. It is argued that for this to work the correlation necessarily has to be transferred from somewhere else. Thus, correlation is a resource that is consumed when measurements take place. It is also argued that a network of such measurements establishes a stable objective classical reality, especially in the context of repeatability of experiments. △ Less

Submitted 26 June, 2023; v1 submitted 7 December, 2022; originally announced December 2022.

Comments: 13 pages, 5 figures

A 3D View of Orion: I. Barnard's Loop

Authors: Michael M. Foley, Alyssa Goodman, Catherine Zucker, John C. Forbes, Ralf Konietzka, Cameren Swiggum, João Alves, John Bally, Juan D. Soler, Josefa E. Großschedl, Shmuel Bialy, Michael Y. Grudić, Reimar Leike, Torsten Ensslin

Abstract: Barnard's Loop is a famous arc of H$αあるふぁ$ emission located in the Orion star-forming region. Here, we provide evidence of a possible formation mechanism for Barnard's Loop and compare our results with recent work suggesting a major feedback event occurred in the region around 6 Myr ago. We present a 3D model of the large-scale Orion region, indicating coherent, radial, 3D expansion of the OBP-Near/Br… ▽ More Barnard's Loop is a famous arc of H$αあるふぁ$ emission located in the Orion star-forming region. Here, we provide evidence of a possible formation mechanism for Barnard's Loop and compare our results with recent work suggesting a major feedback event occurred in the region around 6 Myr ago. We present a 3D model of the large-scale Orion region, indicating coherent, radial, 3D expansion of the OBP-Near/Briceño-1 (OBP-B1) cluster in the middle of a large dust cavity. The large-scale gas in the region also appears to be expanding from a central point, originally proposed to be Orion X. OBP-B1 appears to serve as another possible center, and we evaluate whether Orion X or OBP-B1 is more likely to be the cause of the expansion. We find that neither cluster served as the single expansion center, but rather a combination of feedback from both likely propelled the expansion. Recent 3D dust maps are used to characterize the 3D topology of the entire region, which shows Barnard's Loop's correspondence with a large dust cavity around the OPB-B1 cluster. The molecular clouds Orion A, Orion B, and Orion $λらむだ$ reside on the shell of this cavity. Simple estimates of gravitational effects from both stars and gas indicate that the expansion of this asymmetric cavity likely induced anisotropy in the kinematics of OBP-B1. We conclude that feedback from OBP-B1 has affected the structure of the Orion A, Orion B, and Orion $λらむだ$ molecular clouds and may have played a major role in the formation of Barnard's Loop. △ Less

Submitted 2 December, 2022; originally announced December 2022.

Comments: 25 pages, 10 figures. Submitted to ApJ

Sparse Kernel Gaussian Processes through Iterative Charted Refinement (ICR)

Authors: Gordian Edenhofer, Reimar H. Leike, Philipp Frank, Torsten A. Enßlin

Abstract: Gaussian Processes (GPs) are highly expressive, probabilistic models. A major limitation is their computational complexity. Naively, exact GP inference requires $\mathcal{O}(N^3)$ computations with $N$ denoting the number of modeled points. Current approaches to overcome this limitation either rely on sparse, structured or stochastic representations of data or kernel respectively and usually invol… ▽ More Gaussian Processes (GPs) are highly expressive, probabilistic models. A major limitation is their computational complexity. Naively, exact GP inference requires $\mathcal{O}(N^3)$ computations with $N$ denoting the number of modeled points. Current approaches to overcome this limitation either rely on sparse, structured or stochastic representations of data or kernel respectively and usually involve nested optimizations to evaluate a GP. We present a new, generative method named Iterative Charted Refinement (ICR) to model GPs on nearly arbitrarily spaced points in $\mathcal{O}(N)$ time for decaying kernels without nested optimizations. ICR represents long- as well as short-range correlations by combining views of the modeled locations at varying resolutions with a user-provided coordinate chart. In our experiment with points whose spacings vary over two orders of magnitude, ICR's accuracy is comparable to state-of-the-art GP methods. ICR outperforms existing methods in terms of computational speed by one order of magnitude on the CPU and GPU and has already been successfully applied to model a GP with $122$ billion parameters. △ Less

Submitted 21 June, 2022; originally announced June 2022.

The Galactic 3D large-scale dust distribution via Gaussian process regression on spherical coordinates

Authors: R. H. Leike, G. Edenhofer, J. Knollmüller, C. Alig, P. Frank, T. A. Enßlin

Abstract: Knowing the Galactic 3D dust distribution is relevant for understanding many processes in the interstellar medium and for correcting many astronomical observations for dust absorption and emission. Here, we aim for a 3D reconstruction of the Galactic dust distribution with an increase in the number of meaningful resolution elements by orders of magnitude with respect to previous reconstructions, w… ▽ More Knowing the Galactic 3D dust distribution is relevant for understanding many processes in the interstellar medium and for correcting many astronomical observations for dust absorption and emission. Here, we aim for a 3D reconstruction of the Galactic dust distribution with an increase in the number of meaningful resolution elements by orders of magnitude with respect to previous reconstructions, while taking advantage of the dust's spatial correlations to inform the dust map. We use iterative grid refinement to define a log-normal process in spherical coordinates. This log-normal process assumes a fixed correlation structure, which was inferred in an earlier reconstruction of Galactic dust. Our map is informed through 111 Million data points, combining data of PANSTARRS, 2MASS, Gaia DR2 and ALLWISE. The log-normal process is discretized to 122 Billion degrees of freedom, a factor of 400 more than our previous map. We derive the most probable posterior map and an uncertainty estimate using natural gradient descent and the Fisher-Laplace approximation. The dust reconstruction covers a quarter of the volume of our Galaxy, with a maximum coordinate distance of $16\,\text{kpc}$, and meaningful information can be found up to at distances of $4\,$kpc, still improving upon our earlier map by a factor of 5 in maximal distance, of $900$ in volume, and of about eighteen in angular grid resolution. Unfortunately, the maximum posterior approach chosen to make the reconstruction computational affordable introduces artifacts and reduces the accuracy of our uncertainty estimate. Despite of the apparent limitations of the presented 3D dust map, a good part of the reconstructed structures are confirmed by independent maser observations. Thus, the map is a step towards reliable 3D Galactic cartography and already can serve for a number of tasks, if used with care. △ Less

Submitted 25 April, 2022; originally announced April 2022.

On the Three-Dimensional Structure of Local Molecular Clouds

Authors: Catherine Zucker, Alyssa Goodman, João Alves, Shmuel Bialy, Eric W. Koch, Joshua S. Speagle, Michael M. Foley, Douglas Finkbeiner, Reimar Leike, Torsten Enßlin, Joshua E. G. Peek, Gordian Edenhofer

Abstract: We leverage the 1 pc spatial resolution of the Leike et al. 2020 3D dust map to characterize the three-dimensional structure of nearby molecular clouds ($d \lesssim 400$ pc). We start by "skeletonizing" the clouds in 3D volume density space to determine their "spines," which we project on the sky to constrain cloud distances with $\approx 1\%$ uncertainty. For each cloud, we determine an average r… ▽ More We leverage the 1 pc spatial resolution of the Leike et al. 2020 3D dust map to characterize the three-dimensional structure of nearby molecular clouds ($d \lesssim 400$ pc). We start by "skeletonizing" the clouds in 3D volume density space to determine their "spines," which we project on the sky to constrain cloud distances with $\approx 1\%$ uncertainty. For each cloud, we determine an average radial volume density profile around its 3D spine and fit the profiles using Gaussian and Plummer functions. The radial volume density profiles are well-described by a two-component Gaussian function, consistent with clouds having broad, lower-density outer envelopes and narrow, higher-density inner layers. The ratio of the outer to inner envelope widths is $\approx 3:1$. We hypothesize that these two components may be tracing a transition between atomic and diffuse molecular gas or between the unstable and cold neutral medium. Plummer-like models can also provide a good fit, with molecular clouds exhibiting shallow power-law wings with density, $n$, falling off like $n^{-2}$ at large radii. Using Bayesian model selection, we find that parameterizing the clouds' profiles using a single Gaussian is disfavored. We compare our results with 2D dust extinction maps, finding that the 3D dust recovers the total cloud mass from integrated approaches with fidelity, deviating only at higher levels of extinction ($A_V \gtrsim 2 - 3$ mag). The 3D cloud structure described here will enable comparisons with synthetic clouds generated in simulations, offering unprecedented insight into the origins and fates of molecular clouds in the interstellar medium. △ Less

Submitted 27 September, 2021; v1 submitted 20 September, 2021; originally announced September 2021.

Comments: Published in ApJ. For data behind the figures see https://dataverse.harvard.edu/dataverse/3D_cloud_structure . For interactive gallery of local clouds see https://faun.rc.fas.harvard.edu/czucker/Paper_Figures/3D_Cloud_Topologies/gallery.html . Companion paper to Bialy et al. 2021 (arXiv:2109.09763). For press see https://www.cfa.harvard.edu/news/gigantic-cavity-space-sheds-new-light-how-stars-form

Journal ref: ApJ 919 35 (2021)

The Per-Tau Shell: A Giant Star-Forming Spherical Shell Revealed by 3D Dust Observations

Authors: Shmuel Bialy, Catherine Zucker, Alyssa Goodman, Michael M. Foley, João Alves, Vadim A. Semenov, Robert Benjamin, Reimar Leike, Torsten Enßlin

Abstract: A major question in the field of star formation is how molecular clouds form out of the diffuse Interstellar Medium (ISM). Recent advances in 3D dust mapping are revolutionizing our view of the structure of the ISM. Using the highest-resolution 3D dust map to date, we explore the structure of a nearby star-forming region, which includes the well-known Perseus and Taurus molecular clouds. We reveal… ▽ More A major question in the field of star formation is how molecular clouds form out of the diffuse Interstellar Medium (ISM). Recent advances in 3D dust mapping are revolutionizing our view of the structure of the ISM. Using the highest-resolution 3D dust map to date, we explore the structure of a nearby star-forming region, which includes the well-known Perseus and Taurus molecular clouds. We reveal an extended near-spherical shell, 156 pc in diameter, hereafter the "Per-Tau Shell", in which the Perseus and Taurus clouds are embedded. We also find a large ring structure at the location of Taurus, hereafter, the "Tau Ring". We discuss a formation scenario for the Per-Tau Shell, in which previous stellar and supernova (SN) feedback events formed a large expanding shell, where the swept-up ISM has condensed to form both the shell and the Perseus and Taurus molecular clouds within it. We present auxiliary observations of HI, H$αあるふぁ$, $^{26}$Al, and X-rays that further support this scenario, and estimate Per-Tau Shell's age to be $\approx 6-22$ Myrs. The Per-Tau shell offers the first three-dimensional observational view of a phenomenon long-hypothesized theoretically, molecular cloud formation and star formation triggered by previous stellar and SN feedback. △ Less

Submitted 20 September, 2021; originally announced September 2021.

Comments: ApJ Letters, accepted. Main text: 8 pages, 4 figures. An Interactive + Augmented Reality Figure at https://faun.rc.fas.harvard.edu/czucker/Paper_Figures/sbialy/pertau_superbubble.html

Journal ref: ApJL 919 L5 (2021)

Geometric variational inference

Authors: Philipp Frank, Reimar Leike, Torsten A. Enßlin

Abstract: Efficiently accessing the information contained in non-linear and high dimensional probability distributions remains a core challenge in modern statistics. Traditionally, estimators that go beyond point estimates are either categorized as Variational Inference (VI) or Markov-Chain Monte-Carlo (MCMC) techniques. While MCMC methods that utilize the geometric properties of continuous probability dist… ▽ More Efficiently accessing the information contained in non-linear and high dimensional probability distributions remains a core challenge in modern statistics. Traditionally, estimators that go beyond point estimates are either categorized as Variational Inference (VI) or Markov-Chain Monte-Carlo (MCMC) techniques. While MCMC methods that utilize the geometric properties of continuous probability distributions to increase their efficiency have been proposed, VI methods rarely use the geometry. This work aims to fill this gap and proposes geometric Variational Inference (geoVI), a method based on Riemannian geometry and the Fisher information metric. It is used to construct a coordinate transformation that relates the Riemannian manifold associated with the metric to Euclidean space. The distribution, expressed in the coordinate system induced by the transformation, takes a particularly simple form that allows for an accurate variational approximation by a normal distribution. Furthermore, the algorithmic structure allows for an efficient implementation of geoVI which is demonstrated on multiple examples, ranging from low-dimensional illustrative ones to non-linear, hierarchical Bayesian inverse problems in thousands of dimensions. △ Less

Submitted 2 July, 2021; v1 submitted 21 May, 2021; originally announced May 2021.

Comments: 42 pages, 18 figures, accepted by Entropy

Optical reconstruction of dust in the region of SNR RX J1713.7-3946 from astrometric data

Authors: Reimar Leike, Silvia Celli, Alberto Krone-Martins, Celine Boehm, Martin Glatzle, Yasou Fukui, Hidetoshi Sano, Gavin Rowell

Abstract: The origin of the radiation observed in the region of the supernova remnant RX J1713.7-3946, one of the brightest TeV emitters, has been debated since its discovery. The existence of atomic and molecular clouds in this object supports the idea that part of the GeV gamma rays in this region originate from proton-proton collisions. However, the observed column density of protons derived from gas obs… ▽ More The origin of the radiation observed in the region of the supernova remnant RX J1713.7-3946, one of the brightest TeV emitters, has been debated since its discovery. The existence of atomic and molecular clouds in this object supports the idea that part of the GeV gamma rays in this region originate from proton-proton collisions. However, the observed column density of protons derived from gas observations cannot explain the whole emission. Yet there could be a fraction of protons contained in fainter structures that have note been detected so far. Here we search for faint objects in the line of sight of RX J1713.7-3946 using the principle of light extinction and the ESA/Gaia DR2 astrometric and photometric data. We reveal and locate with precision a number of dust clouds and note that only one appears to be in the vicinity of RX J1713.7-3946. We estimate the embedded mass to $M_{dust} = (7.0 \pm 0.6) \times 10^3 \, M_{\odot}$ which might be big enough to contain the missing protons. Finally, using the fact that the supernova remnant is expected to be located in a dusty environment and that there appears to be only one such structure in the vicinity of RX J1713.7-3946, we set a very precise constrain to the supernova remnant distance, at ($1.12 \pm 0.01$) kpc. △ Less

Submitted 27 January, 2021; v1 submitted 29 November, 2020; originally announced November 2020.

Towards Bayesian Data Compression

Authors: Johannes Harth-Kitzerow, Reimar Leike, Philipp Arras, Torsten A. Enßlin

Abstract: In order to handle large data sets omnipresent in modern science, efficient compression algorithms are necessary. Here, a Bayesian data compression (BDC) algorithm that adapts to the specific measurement situation is derived in the context of signal reconstruction. BDC compresses a data set under conservation of its posterior structure with minimal information loss given the prior knowledge on the… ▽ More In order to handle large data sets omnipresent in modern science, efficient compression algorithms are necessary. Here, a Bayesian data compression (BDC) algorithm that adapts to the specific measurement situation is derived in the context of signal reconstruction. BDC compresses a data set under conservation of its posterior structure with minimal information loss given the prior knowledge on the signal, the quantity of interest. Its basic form is valid for Gaussian priors and likelihoods. For constant noise standard deviation, basic BDC becomes equivalent to a Bayesian analog of principal component analysis. Using Metric Gaussian Variational Inference, BDC generalizes to non-linear settings. In its current form, BDC requires the storage of effective instrument response functions for the compressed data and corresponding noise encoding the posterior covariance structure. Their memory demand counteract the compression gain. In order to improve this, sparsity of the compressed responses can be obtained by separating the data into patches and compressing them separately. The applicability of BDC is demonstrated by applying it to synthetic data and radio astronomical data. Still the algorithm needs further improvement as the computation time of the compression and subsequent inference exceeds the time of the inference with the original data. △ Less

Submitted 29 December, 2020; v1 submitted 20 October, 2020; originally announced October 2020.

Comments: 39 pages, 15 figures, 1 table, for code, see https://gitlab.mpcdf.mpg.de/jharthki/bdc

Journal ref: ANNALEN DER PHYSIK 2021, 2000508

Bayesian decomposition of the Galactic multi-frequency sky using probabilistic autoencoders

Authors: Sara Milosevic, Philipp Frank, Reimar H. Leike, Ancla Müller, Torsten A. Enßlin

Abstract: All-sky observations of the Milky Way show both Galactic and non-Galactic diffuse emission, for example from interstellar matter or the cosmic microwave background (CMB). The different emitters are partly superimposed in the measurements, partly they obscure each other, and sometimes they dominate within a certain spectral range. The decomposition of the underlying radiative components from spectr… ▽ More All-sky observations of the Milky Way show both Galactic and non-Galactic diffuse emission, for example from interstellar matter or the cosmic microwave background (CMB). The different emitters are partly superimposed in the measurements, partly they obscure each other, and sometimes they dominate within a certain spectral range. The decomposition of the underlying radiative components from spectral data is a signal reconstruction problem and often associated with detailed physical modeling and substantial computational effort. We aim to build an effective and self-instructing algorithm detecting the essential spectral information contained Galactic all-sky data covering spectral bands from $γがんま$-ray to radio waves. Utilizing principles from information theory, we develop a state-of-the-art variational autoencoder specialized on the adaption to Gaussian noise statistics. We first derive a generic generative process that leads from a low-dimensional set of emission features to the observed high-dimensional data. We formulate a posterior distribution of these features using Bayesian methods and approximate this posterior with variational inference. The algorithm efficiently encodes the information of 35 Galactic emission data sets in ten latent feature maps. These contain the essential information required to reconstruct the initial data with high fidelity and are ranked by the algorithm according to their significance for data regeneration. The three most significant feature maps encode astrophysical components: (1) The dense interstellar medium (ISM), (2) the hot and dilute regions of the ISM and (3) the CMB. The machine-assisted and data-driven dimensionality reduction of spectral data is able to uncover the physical features encoding the input data. Our algorithm is able to extract the dense and dilute Galactic regions, as well as the CMB, from the sky brightness values only. △ Less

Submitted 14 September, 2020; originally announced September 2020.

Comments: 25 pages, 8 figures, 3 tables. Submitted to Astronomy & Astrophysics

Journal ref: A&A 650, A100 (2021)

Comparison of classical and Bayesian imaging in radio interferometry

Authors: Philipp Arras, Hertzog L. Bester, Richard A. Perley, Reimar Leike, Oleg Smirnov, Rüdiger Westermann, Torsten A. Enßlin

Abstract: CLEAN, the commonly employed imaging algorithm in radio interferometry, suffers from a number of shortcomings: in its basic version it does not have the concept of diffuse flux, and the common practice of convolving the CLEAN components with the CLEAN beam erases the potential for super-resolution; it does not output uncertainty information; it produces images with unphysical negative flux regions… ▽ More CLEAN, the commonly employed imaging algorithm in radio interferometry, suffers from a number of shortcomings: in its basic version it does not have the concept of diffuse flux, and the common practice of convolving the CLEAN components with the CLEAN beam erases the potential for super-resolution; it does not output uncertainty information; it produces images with unphysical negative flux regions; and its results are highly dependent on the so-called weighting scheme as well as on any human choice of CLEAN masks to guiding the imaging. Here, we present the Bayesian imaging algorithm resolve which solves the above problems and naturally leads to super-resolution. We take a VLA observation of Cygnus~A at four different frequencies and image it with single-scale CLEAN, multi-scale CLEAN and resolve. Alongside the sky brightness distribution resolve estimates a baseline-dependent correction function for the noise budget, the Bayesian equivalent of weighting schemes. We report noise correction factors between 0.4 and 429. The enhancements achieved by resolve come at the cost of higher computational effort. △ Less

Submitted 25 January, 2021; v1 submitted 26 August, 2020; originally announced August 2020.

Comments: 23 pages, 16 figures, 4 tables, data published at https://doi.org/10.5281/zenodo.4267057

Journal ref: A&A 646, A84 (2021)

Spatially Resolved Ultraviolet Spectroscopy of the Great Dimming of Betelgeuse

Authors: Andrea K. Dupree, Klaus G. Strassmeier, Lynn D. Matthews, Han Uitenbroek, Thomas Calderwood, Thomas Granzer, Edward F Guinan, Reimar Leike, Miguel Montargès, Anita M. S. Richards, Richard Wasatonic, Michael Weber

Abstract: The bright supergiant, Betelgeuse (Alpha Orionis, HD 39801) experienced a visual dimming during 2019 December and the first quarter of 2020 reaching an historic minimum 2020 February 7$-$13. During 2019 September-November, prior to the optical dimming event, the photosphere was expanding. At the same time, spatially resolved ultraviolet spectra using the Hubble Space Telescope/Space Telescope Imag… ▽ More The bright supergiant, Betelgeuse (Alpha Orionis, HD 39801) experienced a visual dimming during 2019 December and the first quarter of 2020 reaching an historic minimum 2020 February 7$-$13. During 2019 September-November, prior to the optical dimming event, the photosphere was expanding. At the same time, spatially resolved ultraviolet spectra using the Hubble Space Telescope/Space Telescope Imaging Spectrograph revealed a substantial increase in the ultraviolet spectrum and Mg II line emission from the chromosphere over the southern hemisphere of the star. Moreover, the temperature and electron density inferred from the spectrum and C II diagnostics also increased in this hemisphere. These changes happened prior to the Great Dimming Event. Variations in the Mg II k-line profiles suggest material moved outwards in response to the passage of a pulse or acoustic shock from 2019 September through 2019 November. It appears that this extraordinary outflow of material from the star, likely initiated by convective photospheric elements, was enhanced by the coincidence with the outward motions in this phase of the $\sim$400 day pulsation cycle. These ultraviolet observations appear to provide the connecting link between the known large convective cells in the photosphere and the mass ejection event that cooled to form the dust cloud in the southern hemisphere imaged in 2019 December, and led to the exceptional optical dimming of Betelgeuse in 2020 February. △ Less

Submitted 11 August, 2020; originally announced August 2020.

Comments: 11 pages, 8 figures, Astrophysical Journal, accepted

Resolving nearby dust clouds

Authors: R. H. Leike, M. Glatzle, T. A. Enßlin

Abstract: Aims: Mapping the interstellar medium in 3D provides a wealth of insights into its inner working. The Milky Way is the only galaxy for which detailed 3D mapping can be achieved in principle. In this paper, we reconstruct the dust density in and around the local super-bubble. Methods: The combined data from surveys such as Gaia, 2MASS, PANSTARRS, and ALLWISE provide the necessary information to m… ▽ More Aims: Mapping the interstellar medium in 3D provides a wealth of insights into its inner working. The Milky Way is the only galaxy for which detailed 3D mapping can be achieved in principle. In this paper, we reconstruct the dust density in and around the local super-bubble. Methods: The combined data from surveys such as Gaia, 2MASS, PANSTARRS, and ALLWISE provide the necessary information to make detailed maps of the interstellar medium in our surrounding. To this end, we used variational inference and Gaussian processes to model the dust extinction density, exploiting its intrinsic correlations. Results: We reconstructed a highly resolved dust map, showing the nearest dust clouds at a distance of up to 400pc with a resolution of 1pc. Conclusions: Our reconstruction provides insights into the structure of the interstellar medium. We compute summary statistics of the spectral index and the 1-point function of the logarithmic dust extinction density, which may constrain simulations of the interstellar medium that achieve a similar resolution. △ Less

Submitted 5 August, 2020; v1 submitted 14 April, 2020; originally announced April 2020.

Journal ref: A&A 639, A138 (2020)

Variable structures in M87* from space, time and frequency resolved interferometry

Authors: Philipp Arras, Philipp Frank, Philipp Haim, Jakob Knollmüller, Reimar Leike, Martin Reinecke, Torsten Enßlin

Abstract: Observing the dynamics of compact astrophysical objects provides insights into their inner workings, thereby probing physics under extreme conditions. The immediate vicinity of an active supermassive black hole with its event horizon, photon ring, accretion disk, and relativistic jets is a perfect place to study general relativity and magneto-hydrodynamics. The observations of M87* with Very Long… ▽ More Observing the dynamics of compact astrophysical objects provides insights into their inner workings, thereby probing physics under extreme conditions. The immediate vicinity of an active supermassive black hole with its event horizon, photon ring, accretion disk, and relativistic jets is a perfect place to study general relativity and magneto-hydrodynamics. The observations of M87* with Very Long Baseline Interferometry (VLBI) by the Event Horizon Telescope (EHT) allows to investigate its dynamical processes on time scales of days. Compared to regular radio interferometers, VLBI networks typically have fewer antennas and low signal to noise ratios (SNRs). Furthermore, the source is variable, prohibiting integration over time to improve SNR. Here, we present an imaging algorithm that copes with the data scarcity and temporal evolution, while providing uncertainty quantification. Our algorithm views the imaging task as a Bayesian inference problem of a time-varying brightness, exploits the correlation structure in time, and reconstructs a ${2+1+1}$ dimensional time-variable and spectrally resolved image at once. We apply this method to the EHT observation of M87* and validate our approach on synthetic data. The time- and frequency-resolved reconstruction of M87* confirms variable structures on the emission ring. The reconstruction indicates extended and time-variable emission structures outside the ring itself. △ Less

Submitted 5 June, 2022; v1 submitted 12 February, 2020; originally announced February 2020.

Comments: 32 pages, 17 figures, 6 tables

Unified Radio Interferometric Calibration and Imaging with Joint Uncertainty Quantification

Authors: Philipp Arras, Philipp Frank, Reimar Leike, Rüdiger Westermann, Torsten Enßlin

Abstract: The data reduction procedure for radio interferometers can be viewed as a combined calibration and imaging problem. We present an algorithm that unifies cross-calibration, self-calibration, and imaging. Being a Bayesian method, that algorithm does not only calculate an estimate of the sky brightness distribution, but also provides an estimate of the joint uncertainty which entails both the uncerta… ▽ More The data reduction procedure for radio interferometers can be viewed as a combined calibration and imaging problem. We present an algorithm that unifies cross-calibration, self-calibration, and imaging. Being a Bayesian method, that algorithm does not only calculate an estimate of the sky brightness distribution, but also provides an estimate of the joint uncertainty which entails both the uncertainty of the calibration and the one of the actual observation. The algorithm is formulated in the language of information field theory and uses Metric Gaussian Variational Inference (MGVI) as the underlying statistical method. So far only direction-independent antenna-based calibration is considered. This restriction may be released in future work. An implementation of the algorithm is contributed as well. △ Less

Submitted 20 July, 2019; v1 submitted 26 March, 2019; originally announced March 2019.

Comments: 12 pages, 22 figures

Journal ref: A&A 627, A134 (2019)

Field dynamics inference for local and causal interactions

Authors: Philipp Frank, Reimar Leike, Torsten A. Enßlin

Abstract: Inference of fields defined in space and time from observational data is a core discipline in many scientific areas. This work approaches the problem in a Bayesian framework. The proposed method is based on statistically homogeneous random fields defined in space and time and demonstrates how to reconstruct the field together with its prior correlation structure from data. The prior model of the c… ▽ More Inference of fields defined in space and time from observational data is a core discipline in many scientific areas. This work approaches the problem in a Bayesian framework. The proposed method is based on statistically homogeneous random fields defined in space and time and demonstrates how to reconstruct the field together with its prior correlation structure from data. The prior model of the correlation structure is described in a non-parametric fashion and solely builds on fundamental physical assumptions such as space-time homogeneity, locality, and causality. These assumptions are sufficient to successfully infer the field and its prior correlation structure from noisy and incomplete data of a single realization of the process as demonstrated via multiple numerical examples. △ Less

Submitted 4 May, 2021; v1 submitted 5 February, 2019; originally announced February 2019.

Comments: NIFTy5 release paper, 25 pages, 11 figures, code is part of NIFTy5 release at https://gitlab.mpcdf.mpg.de/ift/nifty

Charting nearby dust clouds using Gaia data only

Authors: R. H. Leike, T. A. Enßlin

Abstract: Aims: Highly resolved maps of the local Galactic dust are an important ingredient for sky emission models. In nearly the whole electromagnetic spectrum one can see imprints of dust, many of which originate from dust clouds within 300pc. Having a detailed 3D reconstruction of these local dust clouds enables detailed studies, helps to quantify the impact on other observables and is a milestone neces… ▽ More Aims: Highly resolved maps of the local Galactic dust are an important ingredient for sky emission models. In nearly the whole electromagnetic spectrum one can see imprints of dust, many of which originate from dust clouds within 300pc. Having a detailed 3D reconstruction of these local dust clouds enables detailed studies, helps to quantify the impact on other observables and is a milestone necessary to enable larger reconstructions, as every sightline for more distant objects will pass through the local dust. Methods: To infer the dust density we use parallax and absorption estimates published by the Gaia collaboration in their second data release. We model the dust as a log-normal process using a hierarchical Bayesian model. We also infer non-parametrically the kernel of the log-normal process, which corresponds to the physical spatial correlation power spectrum of the log-density. Results: Using only Gaia data of the second Gaia data release, we reconstruct the 3D dust density and its spatial correlation spectrum in a 600pc cube centered on the Sun. We report a spectral index of the logarithmic dust density of $3.1$ on Fourier scales with wavelengths between 2pc and 125pc. The resulting 3D dust map as well as the power spectrum and posterior samples are publicly available for download. △ Less

Submitted 17 July, 2019; v1 submitted 17 January, 2019; originally announced January 2019.

Journal ref: A&A 631, A32 (2019)

Towards information optimal simulation of partial differential equations

Authors: Reimar H. Leike, Torsten A. Enßlin

Abstract: Most simulation schemes for partial differential equations (PDEs) focus on minimizing a simple error norm of a discretized version of a field. This paper takes a fundamentally different approach; the discretized field is interpreted as data providing information about a real physical field that is unknown. This information is sought to be conserved by the scheme as the field evolves in time. Such… ▽ More Most simulation schemes for partial differential equations (PDEs) focus on minimizing a simple error norm of a discretized version of a field. This paper takes a fundamentally different approach; the discretized field is interpreted as data providing information about a real physical field that is unknown. This information is sought to be conserved by the scheme as the field evolves in time. Such an information theoretic approach to simulation was pursued before by information field dynamics (IFD). In this paper we work out the theory of IFD for nonlinear PDEs in a noiseless Gaussian approximation. The result is an action that can be minimized to obtain an informationally optimal simulation scheme. It can be brought into a closed form using field operators to calculate the appearing Gaussian integrals. The resulting simulation schemes are tested numerically in two instances for the Burgers equation. Their accuracy surpasses finite-difference schemes on the same resolution. The IFD scheme, however, has to be correctly informed on the subgrid correlation structure. In certain limiting cases we recover well-known simulation schemes like spectral Fourier Galerkin methods. We discuss implications of the approximations made. △ Less

Submitted 11 December, 2017; v1 submitted 8 September, 2017; originally announced September 2017.

Journal ref: Phys. Rev. E 97, 033314 (2018)

NIFTy 3 - Numerical Information Field Theory - A Python framework for multicomponent signal inference on HPC clusters

Authors: Theo Steininger, Jait Dixit, Philipp Frank, Maksim Greiner, Sebastian Hutschenreuter, Jakob Knollmüller, Reimar Leike, Natalia Porqueres, Daniel Pumpe, Martin Reinecke, Matevž Šraml, Csongor Varady, Torsten Enßlin

Abstract: NIFTy, "Numerical Information Field Theory", is a software framework designed to ease the development and implementation of field inference algorithms. Field equations are formulated independently of the underlying spatial geometry allowing the user to focus on the algorithmic design. Under the hood, NIFTy ensures that the discretization of the implemented equations is consistent. This enables the… ▽ More NIFTy, "Numerical Information Field Theory", is a software framework designed to ease the development and implementation of field inference algorithms. Field equations are formulated independently of the underlying spatial geometry allowing the user to focus on the algorithmic design. Under the hood, NIFTy ensures that the discretization of the implemented equations is consistent. This enables the user to prototype an algorithm rapidly in 1D and then apply it to high-dimensional real-world problems. This paper introduces NIFTy 3, a major upgrade to the original NIFTy framework. NIFTy 3 allows the user to run inference algorithms on massively parallel high performance computing clusters without changing the implementation of the field equations. It supports n-dimensional Cartesian spaces, spherical spaces, power spaces, and product spaces as well as transforms to their harmonic counterparts. Furthermore, NIFTy 3 is able to treat non-scalar fields. The functionality and performance of the software package is demonstrated with example code, which implements a real inference algorithm from the realm of information field theory. NIFTy 3 is open-source software available under the GNU General Public License v3 (GPL-3) at https://gitlab.mpcdf.mpg.de/ift/NIFTy/ △ Less

Submitted 3 August, 2017; originally announced August 2017.

Comments: 18 pages, 7 figures, 1 table, available at https://gitlab.mpcdf.mpg.de/ift/NIFTy/

Optimal Belief Approximation

Authors: Reimar H. Leike, Torsten A. Enßlin

Abstract: In Bayesian statistics probability distributions express beliefs. However, for many problems the beliefs cannot be computed analytically and approximations of beliefs are needed. We seek a loss function that quantifies how "embarrassing" it is to communicate a given approximation. We reproduce and discuss an old proof showing that there is only one ranking under the requirements that (1) the best… ▽ More In Bayesian statistics probability distributions express beliefs. However, for many problems the beliefs cannot be computed analytically and approximations of beliefs are needed. We seek a loss function that quantifies how "embarrassing" it is to communicate a given approximation. We reproduce and discuss an old proof showing that there is only one ranking under the requirements that (1) the best ranked approximation is the non-approximated belief and (2) that the ranking judges approximations only by their predictions for actual outcomes. The loss function that is obtained in the derivation is equal to the Kullback-Leibler divergence when normalized. This loss function is frequently used in the literature. However, there seems to be confusion about the correct order in which its functional arguments, the approximated and non-approximated beliefs, should be used. The correct order ensures that the recipient of a communication is only deprived of the minimal amount of information. We hope that the elementary derivation settles the apparent confusion. For example when approximating beliefs with Gaussian distributions the optimal approximation is given by moment matching. This is in contrast to many suggested computational schemes. △ Less

Submitted 3 August, 2017; v1 submitted 27 October, 2016; originally announced October 2016.

Comments: made improvements on the proof and the language

Operator Calculus for Information Field Theory

Authors: Reimar H. Leike, Torsten A. Enßlin

Abstract: Signal inference problems with non-Gaussian posteriors can be hard to tackle. Through using the concept of Gibbs free energy these posteriors are rephrased as Gaussian posteriors for the price of computing various expectation values with respect to a Gaussian distribution. We present a new way of translating these expectation values to a language of operators which is similar to that in quantum me… ▽ More Signal inference problems with non-Gaussian posteriors can be hard to tackle. Through using the concept of Gibbs free energy these posteriors are rephrased as Gaussian posteriors for the price of computing various expectation values with respect to a Gaussian distribution. We present a new way of translating these expectation values to a language of operators which is similar to that in quantum mechanics. This simplifies many calculations, for instance such involving log-normal priors. The operator calculus is illustrated by deriving a novel self-calibrating algorithm which is tested with mock data. △ Less

Submitted 21 October, 2016; v1 submitted 2 May, 2016; originally announced May 2016.

Comments: In this version there is a numerical example which demonstrates that the derived algorithm works. The paper was restructured such that it is better readable

Journal ref: PhysRevE. 94.053306 (2016)