Search

Common techniques for measuring refractive indices, such as ellipsometry and goniometry, are ineffective for van der Waals crystal flakes because of their high anisotropy and small, micron-scale, lateral size. To address this, we employ near-field optical microscopy to analyze the guided optical modes within these crystals. By probing these modes in MoS$_2$ flakes with subwavelength spatial resolution at a wavelength of $1570\,\mathrm{nm}$, we determine both the in-plane and out-of-plane permittivity components of MoS$_2$ as $16.11$ and $6.25$, respectively, with a relative uncertainty below $1\%$, while overcoming the limitations of traditional methods.

Deep learning models in medical contexts face challenges like data scarcity, inhomogeneity, and privacy concerns. This study focuses on improving ventricular segmentation in brain MRI images using synthetic data. We employed two latent diffusion models (LDMs): a mask generator trained using 10,000 masks, and a corresponding SPADE image generator optimized using 6,881 scans to create an MRI conditioned on a 3D brain mask. Conditioning the mask generator on ventricular volume in combination with classifier-free guidance enabled the control of the ventricular volume distribution of the generated synthetic images. Next, the performance of the synthetic data was tested using three nnU-Net segmentation models trained on a real, augmented and entirely synthetic data, respectively. The resulting models were tested on a completely independent hold-out dataset of patients with enlarged ventricles, with manual delineation of the ventricles used as ground truth. The model trained on real data showed a mean absolute error (MAE) of 9.09 \pm 12.18 mL in predicted ventricular volume, while the models trained on synthetic and augmented data showed MAEs of 7.52 \pm 4.81 mL and 6.23 \pm 4.33 mL, respectively. Both the synthetic and augmented model also outperformed the state-of-the-art model SynthSeg, which due to limited performance in cases of large ventricular volumes, showed an MAE of 7.73 \pm 12.12 mL with a factor of 3 higher standard deviation. The model trained on augmented data showed the highest Dice score of 0.892 \pm 0.05, slightly outperforming SynthSeg and on par with the model trained on real data. The synthetic model performed similar to SynthSeg. In summary, we provide evidence that guided synthesis of labeled brain MRI data using LDMs improves the segmentation of enlarged ventricles and outperforms existing state-of-the-art segmentation models.

2D materials research has reached significant scientific milestones, accompanied by a rapidly growing industrial sector in the two decades since the field's inception. Such rapid progress requires pushing past the boundary of what is technically and scientifically feasible and carries the risk of disseminating irreproducible research results. This Expert Recommendation addresses the need for enhanced reproducibility in 2D materials science and physics. Through a comprehensive examination of the factors that affect reproducibility the authors present a set of concrete guidelines designed to improve the reliability of research results. The introduction of a Standardised Template for Experimental Procedures (STEP) offers a novel approach to documenting experimental details that are crucial for replication and troubleshooting. We emphasise the importance of involving stakeholders from research, industry, publishing, funding agencies, and policymaking to foster a culture of transparency, reliability, and trust without blind angles and critical oversights. By addressing systemic issues that hinder reproducibility and presenting actionable steps for improvement, we aim to pave the way for more robust research practices in 2D materials science, contributing to the field's scientific maturation and the subsequent development of beneficial technologies., Comment: 24 pages, 3 textboxes, 1 figure, 2 tables

Medical image analysis using deep learning frameworks has advanced healthcare by automating complex tasks, but many existing frameworks lack flexibility, modularity, and user-friendliness. To address these challenges, we introduce Yucca, an open-source AI framework available at https://github.com/Sllambias/yucca, designed specifically for medical imaging applications and built on PyTorch and PyTorch Lightning. Yucca features a three-tiered architecture: Functional, Modules, and Pipeline, providing a comprehensive and customizable solution. Evaluated across diverse tasks such as cerebral microbleeds detection, white matter hyperintensity segmentation, and hippocampus segmentation, Yucca achieves state-of-the-art results, demonstrating its robustness and versatility. Yucca offers a powerful, flexible, and user-friendly platform for medical image analysis, inviting community contributions to advance its capabilities and impact.

Even though novel imaging techniques have been successful in studying brain structure and function, the measured biological signals are often contaminated by multiple sources of noise, arising due to e.g. head movements of the individual being scanned, limited spatial/temporal resolution, or other issues specific to each imaging technology. Data preprocessing (e.g. denoising) is therefore critical. Preprocessing pipelines have become increasingly complex over the years, but also more flexible, and this flexibility can have a significant impact on the final results and conclusions of a given study. This large parameter space is often referred to as multiverse analyses. Here, we provide conceptual and practical tools for statistical analyses that can aggregate multiple pipeline results along with a new sensitivity analysis testing for hypotheses across pipelines such as "no effect across all pipelines" or "at least one pipeline with no effect". The proposed framework is generic and can be applied to any multiverse scenario, but we illustrate its use based on positron emission tomography data.

Using RAW-images in computer vision problems is surprisingly underexplored considering that converting from RAW to RGB does not introduce any new capture information. In this paper, we show that a sufficiently advanced classifier can yield equivalent results on RAW input compared to RGB and present a new public dataset consisting of RAW images and the corresponding converted RGB images. Classifying images directly from RAW is attractive, as it allows for skipping the conversion to RGB, lowering computation time significantly. Two CNN classifiers are used to classify the images in both formats, confirming that classification performance can indeed be preserved. We furthermore show that the total computation time from RAW image data to classification results for RAW images can be up to 8.46 times faster than RGB. These results contribute to the evidence found in related works, that using RAW images as direct input to computer vision algorithms looks very promising., Comment: https://www.kaggle.com/datasets/mathiasviborg/raw-instinct

We investigate the properties of strong (Hb+[OIII]) emitters before and after the end of the Epoch of Reionization from z=8 to z=5.5. We make use of ultra-deep JWST/NIRCam imaging in the Parallel Field of the MIRI Deep Imaging Survey (MIDIS) in the Hubble eXtreme Deep Field (P2-XDF), in order to select prominent (Hb+[OIII]) emitters (with rest EW_0 > 100 Angstroms) at z=5.5-7, based on their flux density enhancement in the F356W band with respect to the spectral energy distribution continuum. We complement our selection with other (Hb+[OIII]) emitters from the literature at similar and higher (z=7-8) redshifts. We find (non-independent) anti-correlations between EW_0(Hb+[OIII]) and both galaxy stellar mass and age, in agreement with previous studies, and a positive correlation with specific star formation rate (sSFR). On the SFR-M* plane, the (Hb+[OIII]) emitters populate both the star-formation main sequence and the starburst region, which become indistinguishable at low stellar masses (log10(M*) < 7.5). We find tentative evidence for a non-monotonic relation between EW_0(Hb+[OIII]) and SFR, such that both parameters correlate with each other at SFR > 1 Msun/yr, while the correlation flattens out at lower SFRs. This suggests that low metallicities producing high EW_0(Hb+[OIII]) could be important at low SFR values. Interestingly, the properties of the strong emitters and other galaxies (33% and 67% of our z=5.5-7 sample, respectively) are similar, including, in many cases, high sSFR. Therefore, it is crucial to consider both emitters and non-emitters to obtain a complete picture of the cosmic star formation activity around the Epoch of Reionization., Comment: 14 pages, 8 figures. Accepted for publication at the ApJ. Version v3 includes minor update to match journal version

General circulation models (GCMs) are the foundation of weather and climate prediction. GCMs are physics-based simulators which combine a numerical solver for large-scale dynamics with tuned representations for small-scale processes such as cloud formation. Recently, machine learning (ML) models trained on reanalysis data achieved comparable or better skill than GCMs for deterministic weather forecasting. However, these models have not demonstrated improved ensemble forecasts, or shown sufficient stability for long-term weather and climate simulations. Here we present the first GCM that combines a differentiable solver for atmospheric dynamics with ML components, and show that it can generate forecasts of deterministic weather, ensemble weather and climate on par with the best ML and physics-based methods. NeuralGCM is competitive with ML models for 1-10 day forecasts, and with the European Centre for Medium-Range Weather Forecasts ensemble prediction for 1-15 day forecasts. With prescribed sea surface temperature, NeuralGCM can accurately track climate metrics such as global mean temperature for multiple decades, and climate forecasts with 140 km resolution exhibit emergent phenomena such as realistic frequency and trajectories of tropical cyclones. For both weather and climate, our approach offers orders of magnitude computational savings over conventional GCMs. Our results show that end-to-end deep learning is compatible with tasks performed by conventional GCMs, and can enhance the large-scale physical simulations that are essential for understanding and predicting the Earth system., Comment: 92 pages, 54 figures. Nature (2024)

The Brain Imaging Data Structure (BIDS) is a community-driven standard for the organization of data and metadata from a growing range of neuroscience modalities. This paper is meant as a history of how the standard has developed and grown over time. We outline the principles behind the project, the mechanisms by which it has been extended, and some of the challenges being addressed as it evolves. We also discuss the lessons learned through the project, with the aim of enabling researchers in other domains to learn from the success of BIDS.

Deep learning-based models in medical imaging often struggle to generalize effectively to new scans due to data heterogeneity arising from differences in hardware, acquisition parameters, population, and artifacts. This limitation presents a significant challenge in adopting machine learning models for clinical practice. We propose an unsupervised method for robust domain adaptation in brain MRI segmentation by leveraging MRI-specific augmentation techniques. To evaluate the effectiveness of our method, we conduct extensive experiments across diverse datasets, modalities, and segmentation tasks, comparing against the state-of-the-art methods. The results show that our proposed approach achieves high accuracy, exhibits broad applicability, and showcases remarkable robustness against domain shift in various tasks, surpassing the state-of-the-art performance in the majority of cases.

Luminous infrared galaxies at high redshifts ($z$>4) include extreme starbursts that build their stellar mass over short periods of time (>100 Myr). These galaxies are considered to be the progenitors of massive quiescent galaxies at intermediate redshifts ($z\sim$2) but their stellar structure and buildup is unknown. Here, we present the first spatially resolved near-infrared imaging of GN20, one of the most luminous dusty star-forming galaxies known to date, observed at an epoch when the Universe was only 1.5 Gyr old. The 5.6$\mu$m image taken with the JWST Mid-Infrared Instrument (MIRI/JWST) shows that GN20 is a very luminous galaxy (M$_\mathrm{1.1\mu m,AB}$=$-$25.01), with a stellar structure composed of a conspicuous central source and an extended envelope. The central source is an unresolved nucleus that carries 9% of the total flux. The nucleus is co-aligned with the peak of the cold dust emission, and offset by 3.9 kpc from the ultraviolet stellar emission. The diffuse stellar envelope is similar in size to the clumpy CO molecular gas distribution. The centroid of the stellar envelope is offset by 1 kpc from the unresolved nucleus, suggesting GN20 is involved in an interaction or merger event supported by its location as the brightest galaxy in a proto-cluster. The stellar size of GN20 is larger by a factor of about 3-5 than known spheroids, disks, and irregulars at $z\sim$4, while its size and low S\'ersic index are similar to those measured in dusty, infrared luminous galaxies at $z\sim$2 of the same mass. GN20 has all the ingredients necessary for evolving into a massive spheroidal quiescent galaxy at intermediate $z$: it is a large, luminous galaxy at $z$=4.05 involved in a short and massive starburst centred in the stellar nucleus and extended over the entire galaxy, out to radii of 4 kpc, and likely induced by the interaction or merger with a member of the proto-cluster., Comment: 7 pages, 4 figures

Clinical needs and technological advances have resulted in increased use of Artificial Intelligence (AI) in clinical decision support. However, such support can introduce new and amplify existing cognitive biases. Through contextual inquiry and interviews, we set out to understand the use of an existing AI support system by ophthalmologists. We identified concerns regarding anchoring bias and a misunderstanding of the AI's capabilities. Following, we evaluated clinicians' perceptions of three bias mitigation strategies as integrated into their existing decision support system. While clinicians recognised the danger of anchoring bias, we identified a concern around the impact of bias mitigation on procedure time. Our participants were divided in their expectations of any positive impact on diagnostic accuracy, stemming from varying reliance on the decision support. Our results provide insights into the challenges of integrating bias mitigation into AI decision support., Comment: CHI '23

A dynamical systems approach to turbulence envisions the flow as a trajectory through a high-dimensional state space transiently visiting the neighbourhoods of unstable simple invariant solutions (E. Hopf, Commun. Appl. Maths 1, 303, 1948). The hope has always been to turn this appealing picture into a predictive framework where the statistics of the flow follows from a weighted sum of the statistics of each simple invariant solution. Two outstanding obstacles have prevented this goal from being achieved: (1) paucity of known solutions and (2) the lack of a rational theory for predicting the required weights. Here we describe a method to substantially solve these problems, and thereby provide the first compelling evidence that the PDFs of a fully developed turbulent flow can be reconstructed with a set of unstable periodic orbits. Our new method for finding solutions uses automatic differentiation, with high-quality guesses constructed by minimising a trajectory-dependent loss function. We use this approach to find hundreds of new solutions in turbulent, two-dimensional Kolmogorov flow. Robust statistical predictions are then computed by learning weights after converting a turbulent trajectory into a Markov chain for which the states are individual solutions, and the nearest solution to a given snapshot is determined using a deep convolutional autoencoder. To our knowledge, this is the first time the PDFs of a spatio-temporally-chaotic system have been successfully reproduced with a set of simple invariant states, and provides a fascinating connection between self-sustaining dynamical processes and the more well-known statistical properties of turbulence.

Aggregated data obtained from job postings provide powerful insights into labor market demands, and emerging skills, and aid job matching. However, most extraction approaches are supervised and thus need costly and time-consuming annotation. To overcome this, we propose Skill Extraction with Weak Supervision. We leverage the European Skills, Competences, Qualifications and Occupations taxonomy to find similar skills in job ads via latent representations. The method shows a strong positive signal, outperforming baselines based on token-level and syntactic patterns., Comment: Accepted in RecSys in HR'22: The 2nd Workshop on Recommender Systems for Human Resources, in conjunction with the 16th ACM Conference on Recommender Systems

Despite their ubiquity throughout science and engineering, only a handful of partial differential equations (PDEs) have analytical, or closed-form solutions. This motivates a vast amount of classical work on numerical simulation of PDEs and more recently, a whirlwind of research into data-driven techniques leveraging machine learning (ML). A recent line of work indicates that a hybrid of classical numerical techniques and machine learning can offer significant improvements over either approach alone. In this work, we show that the choice of the numerical scheme is crucial when incorporating physics-based priors. We build upon Fourier-based spectral methods, which are known to be more efficient than other numerical schemes for simulating PDEs with smooth and periodic solutions. Specifically, we develop ML-augmented spectral solvers for three common PDEs of fluid dynamics. Our models are more accurate (2-4x) than standard spectral solvers at the same resolution but have longer overall runtimes (~2x), due to the additional runtime cost of the neural network component. We also demonstrate a handful of key design principles for combining machine learning and numerical methods for solving PDEs.

Skill Classification (SC) is the task of classifying job competences from job postings. This work is the first in SC applied to Danish job vacancy data. We release the first Danish job posting dataset: Kompetencer (en: competences), annotated for nested spans of competences. To improve upon coarse-grained annotations, we make use of The European Skills, Competences, Qualifications and Occupations (ESCO; le Vrang et al., 2014) taxonomy API to obtain fine-grained labels via distant supervision. We study two setups: The zero-shot and few-shot classification setting. We fine-tune English-based models and RemBERT (Chung et al., 2020) and compare them to in-language Danish models. Our results show RemBERT significantly outperforms all other models in both the zero-shot and the few-shot setting., Comment: 7 pages, accepted to LREC 2022. arXiv admin note: text overlap with arXiv:2204.12811

Skill Extraction (SE) is an important and widely-studied task useful to gain insights into labor market dynamics. However, there is a lacuna of datasets and annotation guidelines; available datasets are few and contain crowd-sourced labels on the span-level or labels from a predefined skill inventory. To address this gap, we introduce SKILLSPAN, a novel SE dataset consisting of 14.5K sentences and over 12.5K annotated spans. We release its respective guidelines created over three different sources annotated for hard and soft skills by domain experts. We introduce a BERT baseline (Devlin et al., 2019). To improve upon this baseline, we experiment with language models that are optimized for long spans (Joshi et al., 2020; Beltagy et al., 2020), continuous pre-training on the job posting domain (Han and Eisenstein, 2019; Gururangan et al., 2020), and multi-task learning (Caruana, 1997). Our results show that the domain-adapted models significantly outperform their non-adapted counterparts, and single-task outperforms multi-task learning., Comment: Accepted to NAACL 2022 Main conference

In this work, we present a switching nonlinear model predictive control (NMPC) algorithm for a dual-hormone artificial pancreas (AP), and we use maximum likelihood estimation (MLE) to identify model parameters. A dual-hormone AP consists of a continuous glucose monitor (CGM), a control algorithm, an insulin pump, and a glucagon pump. The AP is designed with a heuristic to switch between insulin and glucagon as well as state-dependent constraints. We extend an existing glucoregulatory model with glucagon and exercise for simulation, and we use a simpler model for control. We test the AP (NMPC and MLE) using in silico numerical simulations on 50 virtual people with type 1 diabetes. The system is identified for each virtual person based on data generated with the simulation model. The simulations show a mean of 89.3% time in range (3.9-10 mmol/L) and no hypoglycemic events., Comment: In submission, 7 pages, 6 figures

We propose a model-free artificial pancreas (AP) for people with type 1 diabetes. The algorithmic parameters are tuned to a virtual population of 1,000,000 individuals, and the AP repeatedly estimates the basal and bolus insulin requirements necessary for maintaining normal blood glucose levels. Therefore, the AP can be used without healthcare personnel or engineers customizing the algorithm to each user. The estimates are based on bodyweight, measurements from a continuous glucose monitor (CGM), and estimates of the meal carbohydrate contents. In a virtual clinical trial with all 1,000,000 individuals (i.e., a Monte Carlo closed-loop simulation), the AP achieves a mean time in range of more than 87% and almost 89% of the participants satisfy several glycemic targets., Comment: In submission, 6 pages, 7 figures, and 3 tables

Reinforcement learning policies are often represented by neural networks, but programmatic policies are preferred in some cases because they are more interpretable, amenable to formal verification, or generalize better. While efficient algorithms for learning neural policies exist, learning programmatic policies is challenging. Combining imitation-projection and dataset aggregation with a local search heuristic, we present a simple and direct approach to extracting a programmatic policy from a pretrained neural policy. After examining our local search heuristic on a programming by example problem, we demonstrate our programmatic policy extraction method on a pendulum swing-up problem. Both when trained using a hand crafted expert policy and a learned neural policy, our method discovers simple and interpretable policies that perform almost as well as the original.

We will address the problem of determining the existence and asymptotic stability of a non-trivial periodic orbit in dynamical systems described by polynomial vector fields. To this end, we will lean upon the celebrated results of Borg, Olech and Hartman and newer results of Giesl, who all employ the concept of contraction for this purpose. Importantly, we formulate a numerically tractable algebraic test. The developed algorithm is illustrated in a numerical example.