Search

Both transduction and rejection have emerged as important techniques for defending against adversarial perturbations. A recent work by Tram\`er showed that, in the rejection-only case (no transduction), a strong rejection-solution can be turned into a strong (but computationally inefficient) non-rejection solution. This detector-to-classifier reduction has been mostly applied to give evidence that certain claims of strong selective-model solutions are susceptible, leaving the benefits of rejection unclear. On the other hand, a recent work by Goldwasser et al. showed that rejection combined with transduction can give provable guarantees (for certain problems) that cannot be achieved otherwise. Nevertheless, under recent strong adversarial attacks (GMSA, which has been shown to be much more effective than AutoAttack against transduction), Goldwasser et al.'s work was shown to have low performance in a practical deep-learning setting. In this paper, we take a step towards realizing the promise of transduction+rejection in more realistic scenarios. Theoretically, we show that a novel application of Tram\`er's classifier-to-detector technique in the transductive setting can give significantly improved sample-complexity for robust generalization. While our theoretical construction is computationally inefficient, it guides us to identify an efficient transductive algorithm to learn a selective model. Extensive experiments using state of the art attacks (AutoAttack, GMSA) show that our solutions provide significantly better robust accuracy.

The abundance and distribution of ice in protoplanetary disks (PPD) is critical to understand the linkage between the composition of circumstellar matter and the composition of exoplanets. Edge-on PPDs are a useful tool to constrain such ice composition and its location in the disk, as ice spectral signatures can be observed in absorption against the continuum emission arising from the warmer central disk regions. The aim of this work is to model ice absorption features in PPDs and determine how well the abundance of the main ice species across the disk can be determined within the uncertainty of the physical parameter space. The edge-on PPD around HH 48 NE, a target of the JWST ERS program IceAge, is used as a reference system. We use RADMC-3D to raytrace the mid-infrared continuum. Using a constant parameterized ice abundance, ice opacities are added to the dust opacity in regions wherever the disk is cold enough for the main carbon, oxygen and nitrogen carriers to freeze out. The global abundance of the main ice carriers in HH 48 NE can be determined within a factor of 3, when taking the uncertainty of the physical parameters into account. Ice features in PPDs can be saturated at an optical depth <1, due to local saturation. Spatially observed ice optical depths cannot be directly related to column densities due to radiative transfer effects. Vertical snowlines will not be a clear transition due to the radially increasing height of the snowsurface, but their location may be constrained from observations using radiative transfer modeling. Radial snowlines are not really accesible. Not only the ice abundance, but also inclination, settling, grain size distribution and disk mass have strong impact on the observed ice absorption features in disks. Relative changes in ice abundance can be inferred from observations only if the source structure is well constrained, Comment: 18 pages, 12 figures, accepted for publication in Astronomy & Astrophysics

Context. Observations of edge-on disks are an important tool for constraining general protoplanetary disk properties that cannot be determined in any other way. However, most radiative transfer models cannot simultaneously reproduce the spectral energy distributions (SEDs) and resolved scattered light and submillimeter observations of these systems, due to the differences in geometry and dust properties at different wavelengths. Aims. We simultaneously constrain the geometry of the edge-on protoplanetary disk HH 48 NE and the characteristics of the host star. HH 48 NE is part of the JWST early release science program Ice Age. This work serves as a stepping stone towards a better understanding of the disk physical structure and icy chemistry in this particular source. This kind of modeling lays the groundwork for studying other edge-on sources to be observed with the JWST. Methods. We fit a parameterized dust model to HH 48 NE by coupling the radiative transfer code RADMC-3D and an MCMC framework. The dust structure was fitted independently to a compiled SED, a scattered light image at 0.8 ${\mu}$m and an ALMA dust continuum observation at 890 ${\mu}$m. Results. We find that 90% of the dust mass in HH 48 NE is settled to the disk midplane, less than in average disks, and that the atmospheric layers of the disk contain exclusively large grains (0.3-10 ${\mu}$m). The exclusion of small grains in the upper atmosphere likely has important consequences for the chemistry due to the deep penetration of high-energy photons. The addition of a relatively large cavity (ca. 50 au in radius) is necessary to explain the strong mid-infrared emission, and to fit the scattered light and continuum observations simultaneously., Comment: 16 pages, 8 figures, accepted for publication in Astronomy & Astrophysics

The Event Horizon Telescope (EHT) has led to the first images of a supermassive black hole, revealing the central compact objects in the elliptical galaxy M87 and the Milky Way. Proposed upgrades to this array through the next-generation EHT (ngEHT) program would sharply improve the angular resolution, dynamic range, and temporal coverage of the existing EHT observations. These improvements will uniquely enable a wealth of transformative new discoveries related to black hole science, extending from event-horizon-scale studies of strong gravity to studies of explosive transients to the cosmological growth and influence of supermassive black holes. Here, we present the key science goals for the ngEHT and their associated instrument requirements, both of which have been formulated through a multi-year international effort involving hundreds of scientists worldwide., Comment: 32 pages, 11 figures, accepted for publication in a special issue of Galaxies on the ngEHT (https://www.mdpi.com/journal/galaxies/special_issues/ngEHT_blackholes)

ESA's Jupiter Icy Moons Explorer (JUICE) will provide a detailed investigation of the Jovian system in the 2030s, combining a suite of state-of-the-art instruments with an orbital tour tailored to maximise observing opportunities. We review the Jupiter science enabled by the JUICE mission, building on the legacy of discoveries from the Galileo, Cassini, and Juno missions, alongside ground- and space-based observatories. We focus on remote sensing of the climate, meteorology, and chemistry of the atmosphere and auroras from the cloud-forming weather layer, through the upper troposphere, into the stratosphere and ionosphere. The Jupiter orbital tour provides a wealth of opportunities for atmospheric and auroral science: global perspectives with its near-equatorial and inclined phases, sampling all phase angles from dayside to nightside, and investigating phenomena evolving on timescales from minutes to months. The remote sensing payload spans far-UV spectroscopy (50-210 nm), visible imaging (340-1080 nm), visible/near-infrared spectroscopy (0.49-5.56 $\mu$m), and sub-millimetre sounding (near 530-625\,GHz and 1067-1275\,GHz). This is coupled to radio, stellar, and solar occultation opportunities to explore the atmosphere at high vertical resolution; and radio and plasma wave measurements of electric discharges in the Jovian atmosphere and auroras. Cross-disciplinary scientific investigations enable JUICE to explore coupling processes in giant planet atmospheres, to show how the atmosphere is connected to (i) the deep circulation and composition of the hydrogen-dominated interior; and (ii) to the currents and charged particle environments of the external magnetosphere. JUICE will provide a comprehensive characterisation of the atmosphere and auroras of this archetypal giant planet., Comment: 83 pages, 24 figures, accepted to Space Science Reviews special issue on ESA's JUICE mission

The Cichorium genus offers a unique opportunity to study the sporophytic self incompatibility (SSI) system, being composed of species characterized by highly efficient SI (C. intybus) and complete self compatibility (C. endivia). The chicory genome was used to map 7 previously identified SSI locus-associated markers. The region containing the S locus was restricted to an 4 M bp window on chromosome 5. Among the genes predicted in this region, MDIS1 INTERACTING RECEPTOR LIKE KINASE 2 (MIK2) was promising as a candidate for SSI. Its ortholog in Arabidopsis is involved in pollen stigma recognition reactions, and its protein structure is similar to that of S-receptor kinase (SRK), a key component of the SSI in the Brassica genus. The sequencing of MIK2 in chicory and endive accessions revealed two contrasting scenarios. In C. endivia, MIK2 was fully conserved even comparing different botanical varieties (smooth and curly). In C. intybus, 387 SNPs and 3 INDELs were identified when comparing accessions of different biotypes from the same botanical variety (radicchio). The SNP distribution throughout the gene was uneven, with hypervariable domains preferentially localized in the LRR-rich extracellular region, putatively identified as the receptor domain. The gene was hypothesized to be under positive selection, as the nonsynonymous mutations were more than double the synonymous ones (dN / dS = 2.17). An analogous situation was observed analyzing the first 500 bp of the MIK2 promoter: no SNPs were observed among the endive samples, whereas 44 SNPs and 6 INDELs were detected among the chicory samples. Further analyses are needed to confirm the role of MIK2 in SSI and to demonstrate whether the 23 species-specific nonsynonymous SNPs in the CDS and/or the species-specific 10 bp INDEL found in a CCAAT box region of the promoter are responsible for the contrasting sexual behaviors of the two species.

It is well known that noncommutative geometry naturally emerges in the quantum Hall states due to the presence of strong and constant magnetic fields. Here, we discuss the underlying noncommutative geometry of quantum Hall fluids in which the magnetic fields are spatially inhomogenoeus. We analyze these cases by employing symplectic geometry and Fedosov's deformation quantization, which rely on symplectic connections and Fedosov star-product. Through this formalism, we unveil some new features concerning the static limit of the Haldane's unimodular metric and the Girvin-MacDonald-Platzman algebra of the density operators, which plays a central role in the fractional quantum Hall effect., Comment: 6 pages

Interpretation of resolved polarized images of black holes by the Event Horizon Telescope (EHT) requires predictions of the polarized emission observable by an Earth-based instrument for a particular model of the black hole accretion system. Such predictions are generated by general relativistic radiative transfer (GRRT) codes, which integrate the equations of polarized radiative transfer in curved spacetime. A selection of ray-tracing GRRT codes used within the EHT collaboration is evaluated for accuracy and consistency in producing a selection of test images, demonstrating that the various methods and implementations of radiative transfer calculations are highly consistent. When imaging an analytic accretion model, we find that all codes produce images similar within a pixel-wise normalized mean squared error (NMSE) of 0.012 in the worst case. When imaging a snapshot from a cell-based magnetohydrodynamic simulation, we find all test images to be similar within NMSEs of 0.02, 0.04, 0.04, and 0.12 in Stokes I, Q, U , and V respectively. We additionally find the values of several image metrics relevant to published EHT results to be in agreement to much better precision than measurement uncertainties., Comment: Accepted for publication in ApJ

An important goal of online platforms is to enable content discovery, i.e. allow users to find a catalog entity they were not familiar with. A pre-requisite to discover an entity, e.g. a book, with a search engine is that the entity is retrievable, i.e. there are queries for which the system will surface such entity in the top results. However, machine-learned search engines have a high retrievability bias, where the majority of the queries return the same entities. This happens partly due to the predominance of narrow intent queries, where users create queries using the title of an already known entity, e.g. in book search 'harry potter'. The amount of broad queries where users want to discover new entities, e.g. in music search 'chill lyrical electronica with an atmospheric feeling to it', and have a higher tolerance to what they might find, is small in comparison. We focus here on two factors that have a negative impact on the retrievability of the entities (I) the training data used for dense retrieval models and (II) the distribution of narrow and broad intent queries issued in the system. We propose CtrlQGen, a method that generates queries for a chosen underlying intent-narrow or broad. We can use CtrlQGen to improve factor (I) by generating training data for dense retrieval models comprised of diverse synthetic queries. CtrlQGen can also be used to deal with factor (II) by suggesting queries with broader intents to users. Our results on datasets from the domains of music, podcasts, and books reveal that we can significantly decrease the retrievability bias of a dense retrieval model when using CtrlQGen. First, by using the generated queries as training data for dense models we make 9% of the entities retrievable (go from zero to non-zero retrievability). Second, by suggesting broader queries to users, we can make 12% of the entities retrievable in the best case., Comment: Accepted for publication in the International World Wide Web Conference 2023

Contrastive learning is a cornerstone underlying recent progress in multi-view and multimodal learning, e.g., in representation learning with image/caption pairs. While its effectiveness is not yet fully understood, a line of recent work reveals that contrastive learning can invert the data generating process and recover ground truth latent factors shared between views. In this work, we present new identifiability results for multimodal contrastive learning, showing that it is possible to recover shared factors in a more general setup than the multi-view setting studied previously. Specifically, we distinguish between the multi-view setting with one generative mechanism (e.g., multiple cameras of the same type) and the multimodal setting that is characterized by distinct mechanisms (e.g., cameras and microphones). Our work generalizes previous identifiability results by redefining the generative process in terms of distinct mechanisms with modality-specific latent variables. We prove that contrastive learning can block-identify latent factors shared between modalities, even when there are nontrivial dependencies between factors. We empirically verify our identifiability results with numerical simulations and corroborate our findings on a complex multimodal dataset of image/text pairs. Zooming out, our work provides a theoretical basis for multimodal representation learning and explains in which settings multimodal contrastive learning can be effective in practice., Comment: ICLR 2023 camera-ready version