Search

Energy-efficient spintronic devices require a large spin-orbit torque (SOT) and low damping to excite magnetic precession. In conventional devices with heavy-metal/ferromagnet bilayers, reducing the ferromagnet thickness to $\sim$1 nm enhances the SOT but dramatically increases damping. Here, we investigate an alternative approach based on a 10 nm thick single-layer ferromagnet to attain both low damping and a sizable SOT. Instead of relying on a single interface, we continuously break the bulk inversion symmetry with a vertical compositional gradient of two ferromagnetic elements: Fe with low intrinsic damping and Ni with sizable spin-orbit coupling. We find low effective damping parameters of $\alpha_\mathrm{eff} < 5\times10^{-3}$ in the FeNi alloy films, despite the steep compositional gradients. Moreover, we reveal a sizable anti-damping SOT efficiency of $\theta_\mathrm{AD} \approx 0.05$, even without an intentional compositional gradient. Through depth-resolved x-ray diffraction, we identify a lattice strain gradient as crucial symmetry breaking that underpins the SOT. Our findings provide fresh insights into damping and SOTs in single-layer ferromagnets for power-efficient spintronic devices.

We measure the broad impact of galaxy structure on galaxy formation by examining the ongoing star formation and integrated star formation history as revealed through the stellar masses of galaxies at $z < 7$ based on JWST CEERS data from the Extended Groth Strip (EGS). Using the morphological catalog of 3965 visually classified JWST galaxies from Ferreira et al. (2023), we investigate the evolution of stars, and when they form, as a function of morphological type as well as galaxies classified as passive and starburst through spectral energy distributions. Although disk galaxies dominate the structures of galaxies at $z < 7$, we find that these disks are in general either `passive', or on the main-sequence of star formation, and do not contain a large population of starburst galaxies. We also find no significant correlation between morphological type and the star formation rate or colours of galaxies at $z < 7$. In fact, we find that the morphologically classified `spheroids' tend to be blue and are not found to be predominately passive systems at $z > 1.5$. We also find that the stellar mass function for disk galaxies does not evolve significantly during this time, whereas other galaxy types, such as the peculiar population, evolve dramatically, declining at lower redshifts. This indicates that massive peculiars are more common at higher redshifts. We further find that up to $z \sim 7$, the specific star formation rate (sSFR) does not vary with visual morphology, but strongly depends on stellar mass and internal galaxy mass density. This demonstrates that at early epochs galaxy assembly is a mass-driven, rather than a morphologically-driven, process. Quenching of star formation is therefore a mass-dominated process throughout the universe's history, likely due to the presence of supermassive black holes., Comment: MNRAS in press, 20 pages

We present an analysis of rest-frame UV continuum slopes, $\beta$, using a sample of 1011 galaxies at $6.5

We follow up emission line galaxies identified through the near-infrared slitless HST/WFC3 WISP survey with VLT/FORS2 optical spectroscopy. Over 4 WISP fields, we targetted 85 of 138 line emission objects at $0.4

We develop a photometric search method for identifying smouldering galaxies at $5< z < 8$, which are defined to have weak emission lines and thus generally have low specific star formation rates and may even be in a state of (temporary) quiescence. The deep NIRCam imaging (${\sim}29.5$ AB mag, 5$\sigma$) from the JADES second data release is essential for finding these systems, as they are faint, relatively quiescent dwarf galaxies ($M_* \sim 10^{8}$-$10^9$ $\mathrm{M}_\odot)$ in the Epoch of Reionisation (EoR). Moreover, medium-band imaging is key, enabling a clear identification of the lack of emission lines in these galaxies, thus betraying their dormant flame. Owing to the young age of the Universe, combined with the likely bursty star formation in these first dwarf galaxies, conventional colour-selection methods like the UVJ diagram likely miss a large fraction of the quiescent population in the EoR. Indeed, we find that smouldering galaxies constitute a considerable fraction (0.10-0.35) of the EoR dwarf galaxy population ($M_* \sim 10^{8}$-$10^{9}$ $\mathrm{M}_\odot$). As predicted by simulations, these first dwarf galaxies are fragile, the star formation in their shallow potential wells easily snuffed out by feedback-driven winds triggered by secular or merger-driven starbursts, with the smouldering fraction increasing with decreasing stellar mass. Finally, we provide observational constraints on the smouldering galaxy comoving number density (${\sim}10^{-4}$-$10^{-5}$ dex$^{-1}$ Mpc$^{-3}$), which, although hampered by incompleteness, should aid in our understanding of the primordial baryon cycle, as current simulations greatly disagree on whether these systems are rare (${\sim}1\%$) or common (${\sim}50\%$) in the EoR., Comment: 22 pages, 17 figures, 2 tables

While many have shown how Large Language Models (LLMs) can be applied to a diverse set of tasks, the critical issues of data contamination and memorization are often glossed over. In this work, we address this concern for tabular data. Specifically, we introduce a variety of different techniques to assess whether a language model has seen a tabular dataset during training. This investigation reveals that LLMs have memorized many popular tabular datasets verbatim. We then compare the few-shot learning performance of LLMs on datasets that were seen during training to the performance on datasets released after training. We find that LLMs perform better on datasets seen during training, indicating that memorization leads to overfitting. At the same time, LLMs show non-trivial performance on novel datasets and are surprisingly robust to data transformations. We then investigate the in-context statistical learning abilities of LLMs. While LLMs are significantly better than random at solving statistical classification problems, the sample efficiency of few-shot learning lags behind traditional statistical learning algorithms, especially as the dimension of the problem increases. This suggests that much of the observed few-shot performance on novel real-world datasets is due to the LLM's world knowledge. Overall, our results highlight the importance of testing whether an LLM has seen an evaluation dataset during pre-training. We release the https://github.com/interpretml/LLM-Tabular-Memorization-Checker Python package to test LLMs for memorization of tabular datasets., Comment: COLM camera ready

It is well established that stellar discs are destabilized by sharp features in their phase space, driving recurrent spiral modes. We explore the extent to which surface-density breaks in disc galaxies - which represent sharp changes in the gradient of the disc density - drive new spiral modes. We employ linear perturbation theory to investigate how disc breaks alter the eigenmode spectrum of an otherwise pure exponential disc. We find that the presence of a density break gives rise to a set of new, vigorously growing, modes. For a given multiplicity, these edge modes occur in pairs, with closely separated resonances between each pair. The growth rate of edge modes decreases when the break is weakened or moved outward to lower-density regions of the disc. Both down- and up-bending profiles excite edge modes, whose origin can be best understood via the gravitational torques they exert on the underlying disc. When the profile is down-bending (Type II) the faster growing mode is the inner one while in the up-bending (Type III) case the outer mode is faster growing. In both cases the faster growing mode has a corotation almost coincident with the break. We show that the torques of the edge modes tend to smoothen the break.

Scalar-tensor theories have taken on a key role in attempts to confront the growing open questions in standard cosmology. It is important to understand entirely their dynamics at perturbative level including any possible spatial dependence in their growth of large scale structures. In this work, we investigate the spatial dependence of the growth rate of scalar-tensor theories through the M\'{e}sz\'{a}ros equation. We confirm that at subhorizon level this dependence does not play a major role for viable models. However, we establish conditions on which this criterion is met which may be important for developing new models. In our work, we consider three specific models that exhibit spatial dependence of the growth rate at subhorizon modes, which may also be important for early Universe models.

While many have shown how Large Language Models (LLMs) can be applied to a diverse set of tasks, the critical issues of data contamination and memorization are often glossed over. In this work, we address this concern for tabular data. Starting with simple qualitative tests for whether an LLM knows the names and values of features, we introduce a variety of different techniques to assess the degrees of contamination, including statistical tests for conditional distribution modeling and four tests that identify memorization. Our investigation reveals that LLMs are pre-trained on many popular tabular datasets. This exposure can lead to invalid performance evaluation on downstream tasks because the LLMs have, in effect, been fit to the test set. Interestingly, we also identify a regime where the language model reproduces important statistics of the data, but fails to reproduce the dataset verbatim. On these datasets, although seen during training, good performance on downstream tasks might not be due to overfitting. Our findings underscore the need for ensuring data integrity in machine learning tasks with LLMs. To facilitate future research, we release an open-source tool that can perform various tests for memorization \url{https://github.com/interpretml/LLM-Tabular-Memorization-Checker}., Comment: Table Representation Learning Workshop at NeurIPS 2023

We utilize deep JWST NIRCam observations for the first direct constraints on the Galaxy Stellar Mass Function (GSMF) at z>10. Our EPOCHS v1 sample includes 1120 galaxy candidates at 6.5 9 than some studies predict, suggesting a rapid buildup of stellar mass in the early Universe. We find no incompatibility between our results and those of standard cosmological models, as suggested previously, although the most massive galaxies may require a high star formation efficiency. We find that the 'Little Red Dot' galaxies dominate the z=7 GSMF at high-masses, necessitating a better understanding of the relative contributions of AGN and stellar emission. We show that assuming a theoretically motivated top-heavy IMF reduces stellar mass by 0.5 dex without affecting fit quality, but our results remain consistent with existing cosmological models with a standard IMF., Comment: 40 pages, 15 figures. Submitted to ApJ. Data is avaible on Github: https://github.com/tHarvey303/EpochsIV. Comments to corresponding author welcome at thomas.harvey-3@manchester.ac.uk

Energy efficiency of electronic digital processors is primarily limited by the energy consumption of electronic communication and interconnects. The industry is almost unanimously pushing towards replacing both long-haul, as well as local chip interconnects, using optics to drastically increase efficiency. In this paper, we explore what comes after the successful migration to optical interconnects, as with this inefficiency solved, the main source of energy consumption will be electronic digital computing, memory and electro-optical conversion. Our approach attempts to address all these issues by introducing efficient all-optical digital computing and memory, which in turn eliminates the need for electro-optical conversions. Here, we demonstrate for the first time a scheme to enable general purpose digital data processing in an integrated form and present our photonic integrated circuit (PIC) implementation. For this demonstration we implemented a URISC architecture capable of running any classical piece of software all-optically and present a comprehensive architectural framework for all-optical computing to go beyond., Comment: 14 pages, 10 figures

Recent years have seen important advances in the building of interpretable models, machine learning models that are designed to be easily understood by humans. In this work, we show that large language models (LLMs) are remarkably good at working with interpretable models, too. In particular, we show that LLMs can describe, interpret, and debug Generalized Additive Models (GAMs). Combining the flexibility of LLMs with the breadth of statistical patterns accurately described by GAMs enables dataset summarization, question answering, and model critique. LLMs can also improve the interaction between domain experts and interpretable models, and generate hypotheses about the underlying phenomenon. We release \url{https://github.com/interpretml/TalkToEBM} as an open-source LLM-GAM interface., Comment: XAI4Sci Workshop at AAAI-24

Interpretable machine learning has exploded as an area of interest over the last decade, sparked by the rise of increasingly large datasets and deep neural networks. Simultaneously, large language models (LLMs) have demonstrated remarkable capabilities across a wide array of tasks, offering a chance to rethink opportunities in interpretable machine learning. Notably, the capability to explain in natural language allows LLMs to expand the scale and complexity of patterns that can be given to a human. However, these new capabilities raise new challenges, such as hallucinated explanations and immense computational costs. In this position paper, we start by reviewing existing methods to evaluate the emerging field of LLM interpretation (both interpreting LLMs and using LLMs for explanation). We contend that, despite their limitations, LLMs hold the opportunity to redefine interpretability with a more ambitious scope across many applications, including in auditing LLMs themselves. We highlight two emerging research priorities for LLM interpretation: using LLMs to directly analyze new datasets and to generate interactive explanations., Comment: 7 pages

Electrical manipulation of magnetic order by current-induced spin torques lays the foundation for spintronics. One promising approach is encoding information in the N\'eel vector of antiferromagnetic (AFM) materials, particularly to collinear antiferromagnets with the perpendicular magnetic anisotropy (PMA), as the negligible stray fields and terahertz spin dynamics can enable memory devices with higher integration density and ultrafast speed. Here we demonstrate that the N\'eel order information in a prototypical collinear AFM insulator with PMA, Cr2O3, can be reliably readout via the anomalous Hall effect and efficiently switched by the spin-orbit torque (SOT) effect with a low current density of 5.8*106 A/cm2. Moreover, using Cr2O3 as a mediator, we electrically switch the magnetization of a Y3Fe5O12 film exchange-coupled to the Cr2O3 layer, unambiguously confirming the N\'eel order switching of the Cr2O3 layer. This work provides a significant basis for developing AFM memory devices based on collinear AFM materials with PMA.

Explainability has become a valuable tool in the last few years, helping humans better understand AI-guided decisions. However, the classic explainability tools are sometimes quite limited when considering high-dimensional inputs and neural network classifiers. We present a new explainability method using theoretically proven high-dimensional properties in neural network classifiers. We present two usages of it: 1) On the classical sentiment analysis task for the IMDB reviews dataset, and 2) our Malware-Detection task for our PowerShell scripts dataset., Comment: Accepted to "XAI in Action" workshop @ NeurIPS 2023

We present an analysis of the quenching of star formation in massive galaxies ($M_* > 10^{9.5} M_\odot$) within the first 0.5 - 3 Gyr of the Universe's history utilizing JWST-CEERS data. We utilize a combination of advanced statistical methods to accurately constrain the intrinsic dependence of quenching in a multi-dimensional and inter-correlated parameter space. Specifically, we apply Random Forest (RF) classification, area statistics, and a partial correlation analysis to the JWST-CEERS data. First, we identify the key testable predictions from two state-of-the-art cosmological simulations (IllustrisTNG & EAGLE). Both simulations predict that quenching should be regulated by supermassive black hole mass in the early Universe. Furthermore, both simulations identify the stellar potential ($\phi_*$) as the optimal proxy for black hole mass in photometric data. In photometric observations, where we have no direct constraints on black hole masses, we find that the stellar potential is the most predictive parameter of massive galaxy quenching at all epochs from $z = 0 - 8$, exactly as predicted by simulations for this sample. The stellar potential outperforms stellar mass, galaxy size, galaxy density, and S\'ersic index as a predictor of quiescence at all epochs probed in JWST-CEERS. Collectively, these results strongly imply a stable quenching mechanism operating throughout cosmic history, which is closely connected to the central gravitational potential in galaxies. This connection is explained in cosmological models via massive black holes forming and growing in deep potential wells, and subsequently quenching galaxies through a mix of ejective and preventative active galactic nucleus (AGN) feedback., Comment: Accepted to ApJ. 32 pages; 10 figures

Although most pregnancies result in a good outcome, complications are not uncommon and can be associated with serious implications for mothers and babies. Predictive modeling has the potential to improve outcomes through better understanding of risk factors, heightened surveillance for high risk patients, and more timely and appropriate interventions, thereby helping obstetricians deliver better care. We identify and study the most important risk factors for four types of pregnancy complications: (i) severe maternal morbidity, (ii) shoulder dystocia, (iii) preterm preeclampsia, and (iv) antepartum stillbirth. We use an Explainable Boosting Machine (EBM), a high-accuracy glass-box learning method, for prediction and identification of important risk factors. We undertake external validation and perform an extensive robustness analysis of the EBM models. EBMs match the accuracy of other black-box ML methods such as deep neural networks and random forests, and outperform logistic regression, while being more interpretable. EBMs prove to be robust. The interpretability of the EBM models reveals surprising insights into the features contributing to risk (e.g. maternal height is the second most important feature for shoulder dystocia) and may have potential for clinical application in the prediction and prevention of serious complications in pregnancy., Comment: 25 pages (including appendix and references), 12 figures, 2 tables. J Healthc Inform Res (2023)

The distortion of corner-sharing octahedra in isovalent perovskite transition-metal oxide interfaces is proven to be an excellent way to tailor the electronic and magnetic properties of their heterostructures. Combining depth-dependent magnetic characterization technique; (polarized neutron reflectivity, PNR); and theoretical calculation (density functional theory), we report interface-driven magnetic exchange interactions due to a modification in the octahedral rotations at the interfaces in an isovalent La0.67Ca0.33MnO3 (LCMO)/La0.67Sr0.33MnO3 (LSMO) heterostructures. PNR results determined a length scale of ~ 8 unit cells at the interface, which demonstrated a modification in magnetic properties. The results also predicted a low-temperature exchange bias for these ferromagnetic heterostructures with a maximum exchange bias for the heterostructure, which showed an enhanced antiferromagnetic coupling at the interfaces.

We present the analysis of the stellar population and star formation history of 181 MIRI selected galaxies at redshift 0-3.5 in the massive galaxy cluster field SMACS J0723.3-7327, commonly referred to as SMACS0723, using the James Webb Space Telescope (JWST) Mid-Infrared Instrument (MIRI). We combine the data with the JWST Near Infrared Camera (NIRCam) catalogue, in conjunction with the Hubble Space Telescope (HST) WFC3/IR and ACS imaging. We find that the MIRI bands capture PAH features and dust emission, significantly enhancing the accuracy of photometric redshift and measurements of the physical properties of these galaxies. The median photo-z's of galaxies with MIRI data are found to have a small 0.1% difference from spectroscopic redshifts and reducing the error by 20 percent. With MIRI data included in SED fits, we find that the measured stellar masses are unchanged, while the star formation rate is systematically lower by 0.1 dex. We also fit the median SED of active galactic nuclei (AGN) and star forming galaxies (SFG) separately. MIRI data provides tighter constraints on the AGN contribution, reducing the typical AGN contributions by ~14 percent. In addition, we also compare the median SED obtained with and without MIRI, and we find that including MIRI data yields steeper optical and UV slopes, indicating bluer colours, lower dust attenuation, and younger stellar populations. In the future, MIRI/MRS will enhance our understanding by providing more detailed spectral information and allowing for the study of specific emission features and diagnostics associated with AGN., Comment: 15 pages, 14 figures and 1 table, submitted to MNRAS

We present the results of a size and structural analysis of 1395 galaxies at $0.5 \leq z \lesssim 8$ with stellar masses $\log \left(M_* / M_{\odot}\right)$ $>$ 9.5 within the JWST Public CEERS field that overlaps with the HST CANDELS EGS observations. We use GALFIT to fit single S\'ersic models to the rest-frame optical profile of our galaxies, which is a mass-selected sample complete to our redshift and mass limit. Our primary result is that at fixed rest-frame wavelength and stellar mass, galaxies get progressively smaller, evolving as $\sim (1+z)^{-0.71\pm0.19}$ up to $z \sim 8$. We discover that the vast majority of massive galaxies at high redshifts have low S\'ersic indices, thus do not contain steep, concentrated light profiles. Additionally, we explore the evolution of the size-stellar mass relationship, finding a correlation such that more massive systems are larger up to $z \sim 3$. This relationship breaks down at $z > 3$, where we find that galaxies are of similar sizes, regardless of their star formation rates and S\'ersic index, varying little with mass. We show that galaxies are more compact at redder wavelengths, independent of sSFR or stellar mass up to $z \sim 3$. We demonstrate the size evolution of galaxies continues up to $z \sim 8$, showing that the process or causes for this evolution is active at early times. We discuss these results in terms of ideas behind galaxy formation and evolution at early epochs, such as their importance in tracing processes driving size evolution, including minor mergers and AGN activity., Comment: Submitted to MNRAS

Treatment pathways are step-by-step plans outlining the recommended medical care for specific diseases; they get revised when different treatments are found to improve patient outcomes. Examining health records is an important part of this revision process, but inferring patients' actual treatments from health data is challenging due to complex event-coding schemes and the absence of pathway-related annotations. This study aims to infer the actual treatment steps for a particular patient group from administrative health records (AHR) - a common form of tabular healthcare data - and address several technique- and methodology-based gaps in treatment pathway-inference research. We introduce Defrag, a method for examining AHRs to infer the real-world treatment steps for a particular patient group. Defrag learns the semantic and temporal meaning of healthcare event sequences, allowing it to reliably infer treatment steps from complex healthcare data. To our knowledge, Defrag is the first pathway-inference method to utilise a neural network (NN), an approach made possible by a novel, self-supervised learning objective. We also developed a testing and validation framework for pathway inference, which we use to characterise and evaluate Defrag's pathway inference ability and compare against baselines. We demonstrate Defrag's effectiveness by identifying best-practice pathway fragments for breast cancer, lung cancer, and melanoma in public healthcare records. Additionally, we use synthetic data experiments to demonstrate the characteristics of the Defrag method, and to compare Defrag to several baselines where it significantly outperforms non-NN-based methods. Defrag significantly outperforms several existing pathway-inference methods and offers an innovative and effective approach for inferring treatment pathways from AHRs. Open-source code is provided to encourage further research in this area.

Physical evolution of cosmological models can be tested by using expansion data, while growth history of these models is capable of testing dynamics of the inhomogeneous parts of energy density. The growth factor, as well as its growth index, gives a clear indication of the performance of cosmological models in the regime of structure formation of early Universe. In this work, we explore the growth index in several leading $f(T)$ cosmological models, based on a specific class of teleparallel gravity theories. These have become prominent in the literature and lead to other formulations of teleparallel gravity. Here we adopt a generalized approach by obtaining the M\'{e}sz\'{a}ros equation without immediately imposing the subhorizon limit, because this assumption could lead to over-simplification. This approach gives avenue to study at which $k$ modes the subhorizon limit starts to apply. We obtain numerical results for the growth factor and growth index for a variety of data set combinations for each $f(T)$ model.

Machine learning provides many powerful and effective techniques for analysing heterogeneous electronic health records (EHR). Administrative Health Records (AHR) are a subset of EHR collected for administrative purposes, and the use of machine learning on AHRs is a growing subfield of EHR analytics. Existing reviews of EHR analytics emphasise that the data-modality of the EHR limits the breadth of suitable machine learning techniques, and pursuable healthcare applications. Despite emphasising the importance of data modality, the literature fails to analyse which techniques and applications are relevant to AHRs. AHRs contain uniquely well-structured, categorically encoded records which are distinct from other data-modalities captured by EHRs, and they can provide valuable information pertaining to how patients interact with the healthcare system. This paper systematically reviews AHR-based research, analysing 70 relevant studies and spanning multiple databases. We identify and analyse which machine learning techniques are applied to AHRs and which health informatics applications are pursued in AHR-based research. We also analyse how these techniques are applied in pursuit of each application, and identify the limitations of these approaches. We find that while AHR-based studies are disconnected from each other, the use of AHRs in health informatics research is substantial and accelerating. Our synthesis of these studies highlights the utility of AHRs for pursuing increasingly complex and diverse research objectives despite a number of pervading data- and technique-based limitations. Finally, through our findings, we propose a set of future research directions that can enhance the utility of AHR data and machine learning techniques for health informatics research.

We show that large language models (LLMs) are remarkably good at working with interpretable models that decompose complex outcomes into univariate graph-represented components. By adopting a hierarchical approach to reasoning, LLMs can provide comprehensive model-level summaries without ever requiring the entire model to fit in context. This approach enables LLMs to apply their extensive background knowledge to automate common tasks in data science such as detecting anomalies that contradict prior knowledge, describing potential reasons for the anomalies, and suggesting repairs that would remove the anomalies. We use multiple examples in healthcare to demonstrate the utility of these new capabilities of LLMs, with particular emphasis on Generalized Additive Models (GAMs). Finally, we present the package $\texttt{TalkToEBM}$ as an open-source LLM-GAM interface.

We consider a patient risk models which has access to patient features such as vital signs, lab values, and prior history but does not have access to a patient's diagnosis. For example, this occurs in a model deployed at intake time for triage purposes. We show that such `all-cause' risk models have good generalization across diagnoses but have a predictable failure mode. When the same lab/vital/history profiles can result from diagnoses with different risk profiles (e.g. E.coli vs. MRSA) the risk estimate is a probability weighted average of these two profiles. This leads to an under-estimation of risk for rare but highly risky diagnoses. We propose a fix for this problem by explicitly modeling the uncertainty in risk prediction coming from uncertainty in patient diagnoses. This gives practitioners an interpretable way to understand patient risk beyond a single risk number.

With the successful launch and commissioning of JWST we are now able to routinely spectroscopically probe the rest-frame optical emission of galaxies at $z>6$ for the first time. Amongst the most useful spectral diagnostics used in the optical is the Balmer/4000~\AA\ break; this is, in principle, a diagnostic of the mean ages of composite stellar populations. However, the Balmer break is also sensitive to the shape of the star formation history, the stellar (and gas) metallicity, the presence of nebular continuum emission, and dust attenuation. In this work we explore the origin of the Balmer/4000~\AA\ break using the SYNTHESIZER synthetic observations package. We then make predictions of the Balmer/4000~\AA\ break using the First Light and Reionisation Epoch Simulations (FLARES) at $5

As the deployment of computer vision technology becomes increasingly common in applications of consequence such as medicine or science, the need for explanations of the system output has become a focus of great concern. Unfortunately, many state-of-the-art computer vision models are opaque, making their use challenging from an explanation standpoint, and current approaches to explaining these opaque models have stark limitations and have been the subject of serious criticism. In contrast, Explainable Boosting Machines (EBMs) are a class of models that are easy to interpret and achieve performance on par with the very best-performing models, however, to date EBMs have been limited solely to tabular data. Driven by the pressing need for interpretable models in science, we propose the use of EBMs for scientific image data. Inspired by an important application underpinning the development of quantum technologies, we apply EBMs to cold-atom soliton image data, and, in doing so, demonstrate EBMs for image data for the first time. To tabularize the image data we employ Gabor Wavelet Transform-based techniques that preserve the spatial structure of the data. We show that our approach provides better explanations than other state-of-the-art explainability methods for images., Comment: 12 pages, 2 figures

Missing values are a fundamental problem in data science. Many datasets have missing values that must be properly handled because the way missing values are treated can have large impact on the resulting machine learning model. In medical applications, the consequences may affect healthcare decisions. There are many methods in the literature for dealing with missing values, including state-of-the-art methods which often depend on black-box models for imputation. In this work, we show how recent advances in interpretable machine learning provide a new perspective for understanding and tackling the missing value problem. We propose methods based on high-accuracy glass-box Explainable Boosting Machines (EBMs) that can help users (1) gain new insights on missingness mechanisms and better understand the causes of missingness, and (2) detect -- or even alleviate -- potential risks introduced by imputation algorithms. Experiments on real-world medical datasets illustrate the effectiveness of the proposed methods., Comment: Preprint of a paper accepted by CHIL 2023

Machine learning (ML) recourse techniques are increasingly used in high-stakes domains, providing end users with actions to alter ML predictions, but they assume ML developers understand what input variables can be changed. However, a recourse plan's actionability is subjective and unlikely to match developers' expectations completely. We present GAM Coach, a novel open-source system that adapts integer linear programming to generate customizable counterfactual explanations for Generalized Additive Models (GAMs), and leverages interactive visualizations to enable end users to iteratively generate recourse plans meeting their needs. A quantitative user study with 41 participants shows our tool is usable and useful, and users prefer personalized recourse plans over generic plans. Through a log analysis, we explore how users discover satisfactory recourse plans, and provide empirical evidence that transparency can lead to more opportunities for everyday users to discover counterintuitive patterns in ML models. GAM Coach is available at: https://poloclub.github.io/gam-coach/., Comment: Accepted to CHI 2023. 20 pages, 12 figures. For a demo video, see https://youtu.be/ubacP34H9XE. For a live demo, visit https://poloclub.github.io/gam-coach/

From the TNG50 cosmological simulation we build a sample of 191 well-resolved barred galaxies with a stellar mass $\logMstar > 10$ at $z=0$. We search for box/peanut bulges (BPs) in this sample, finding them in 55\% of cases. We compute $\fbp$, the BP probability for barred galaxies as a function of $\Mstar$, and find that this rises to a plateau, as found in observations of nearby galaxies. The transition mass where $\fbp$ reaches half the plateau value is $\logMstar = 10.14$, consistent with the observational value within measurement errors. We show that this transition in $\fbp$ can be attributed to the youth of the bars at low $\Mstar$, which is a consequence of downsizing of galaxies. Young bars, being generally shorter and weaker, have not yet had time to form BPs. At high mass, while we find a plateau, the value is at $\sim 60\%$ whereas observations saturate at $100\%$. We attribute this difference to excessive heating in TNG50, due to merger activity and to numerical resolution effects. BPs in TNG50 tend to occur in galaxies with more quiescent merger histories. As a result, the main driver of whether a bar hosts a BP in TNG50 is not the galaxy mass, but how long and strong the bar is. Separating the BP sample into those that have visibly buckled and those that have not, we find that fully half of BP galaxies show clear signs of buckling, despite the excessive heating and limited vertical resolution of TNG50., Comment: 20 pages, 24 figures, 5 tables. Submitted to MNRAS

JWST has now made it possible to probe the rest-frame optical line emission of high-redshift galaxies extending to z~9, and potentially beyond. To aid in the interpretation of these emerging constraints, in this work we explore predictions for [OIII] emission in high-redshift galaxies using the First Light and Reionisation Epoch Simulations (FLARES). We produce predictions for the [OIII] luminosity function, its correlation with the UV luminosity, and the distribution of equivalent widths (EWs). We also explore how the [OIII] EW correlates with physical properties including specific star formation rate, metallicity, and dust attenuation. Our predictions are largely consistent with recent observational constraints on the luminosity function, average equivalent widths, and line ratios. However, they fail to reproduce the observed tail of high-EW sources and the number density of extreme line emitters. Possibilities to explain these discrepancies include an additional source of ionising photons and/or greater stochasticity in star formation in the model or photometric scatter and/or bias in the observations. With JWST now rapidly building larger samples and a wider range of emission lines the answer to this remaining discrepancy should be available imminently., Comment: 15 pages, accepted for publication in MNRAS, minor changes from original version

Monitoring the activity of the heart is important for diagnosing and preventing cardiovascular diseases. The electrocardiogram (ECG) is the gold standard for diagnosing such diseases. It monitors the heart's electrical activity, and while this is highly correlated with the cardiac mechanical activity, it does not provide all the information. Other sensors such as the echocardiograph are able to monitor the heart's movements, but such tools are expensive and hard to operate. Therefore, contactless monitoring of the heart using RF sensing has gained interest in recent years. In this paper, we describe a process to extract the movements of the heart from millimeter wave radar with high accuracy, and thus we provide a noninvasive and affordable way to monitor cardiac movements. We then demonstrate the correlation between the observed movements and the ECG. Furthermore, we propose an algorithm to synchronize the ECG signal and the processed signal from the radar sensor. The results we obtained provide insights on the mechanical activity of the heart, which could assist cardiologists in their diagnoses., Comment: 12 pages, 24 figures

Teleparallel geometry offers a platform on which to build up theories of gravity where torsion rather than curvature mediates gravitational interaction. The teleparallel analogue of Horndeski gravity is an approach to teleparallel geometry where scalar-tensor theories are considered in this torsional framework. Being teleparallel gravity of lower order in dynamics, this turns out to be more general than metric Horndeski gravity. In other words, the class of teleparallel Horndeski gravity models is much broader than the standard metric one. In this work, we explore constraints on this wide range of models coming from ghost and Laplacian instabilities. The aim is to limit pathological branches of the theory by fundamental considerations. It is possible to conclude that a very large class of models results physically viable.