Search

As artificial intelligence systems become increasingly integrated into daily life, the field of explainability has gained significant attention. This trend is particularly driven by the complexity of modern AI models and their decision-making processes. The advent of foundation models, characterized by their extensive generalization capabilities and emergent uses, has further complicated this landscape. Foundation models occupy an ambiguous position in the explainability domain: their complexity makes them inherently challenging to interpret, yet they are increasingly leveraged as tools to construct explainable models. In this survey, we explore the intersection of foundation models and eXplainable AI (XAI) in the vision domain. We begin by compiling a comprehensive corpus of papers that bridge these fields. Next, we categorize these works based on their architectural characteristics. We then discuss the challenges faced by current research in integrating XAI within foundation models. Furthermore, we review common evaluation methodologies for these combined approaches. Finally, we present key observations and insights from our survey, offering directions for future research in this rapidly evolving field.

Gallium oxide is a novel advanced material gaining increasing attention for its unique combination of functional properties. It forms in several phases or polymorphs - {\alpha}, {\beta}, {\gamma}, and {\kappa} - having variable properties because of different lattice symmetries. Optical properties are of particular importance as they determine specific device applications and can also be used for phase identification. However, a direct comparison of optical polymorph signatures, including such critical parameters as bandgaps, is challenging due to the scattered, limited, or even absent data for certain phases in the literature. To address this issue, in the present work we systematically cross-correlate optical emission and absorption features of {\alpha}, {\beta}, {\gamma}, and {\kappa} thin films, as well as differently oriented {\beta}-phase bulk crystals and {\gamma}/{\beta} double polymorph structures. We demonstrate that the optical bandgap and emission features scale consistently across these polymorphs upon minimization of the methodological uncertainties. As a result, this work provides a comparative library of near- and far-field optical signatures of the polymorphs for use by the multidisciplinary research community working with gallium oxide., Comment: 17 pages, 14 figures, 4 tables

In this paper, we introduce PASTA (Perceptual Assessment System for explanaTion of Artificial intelligence), a novel framework for a human-centric evaluation of XAI techniques in computer vision. Our first key contribution is a human evaluation of XAI explanations on four diverse datasets (COCO, Pascal Parts, Cats Dogs Cars, and MonumAI) which constitutes the first large-scale benchmark dataset for XAI, with annotations at both the image and concept levels. This dataset allows for robust evaluation and comparison across various XAI methods. Our second major contribution is a data-based metric for assessing the interpretability of explanations. It mimics human preferences, based on a database of human evaluations of explanations in the PASTA-dataset. With its dataset and metric, the PASTA framework provides consistent and reliable comparisons between XAI techniques, in a way that is scalable but still aligned with human evaluations. Additionally, our benchmark allows for comparisons between explanations across different modalities, an aspect previously unaddressed. Our findings indicate that humans tend to prefer saliency maps over other explanation types. Moreover, we provide evidence that human assessments show a low correlation with existing XAI metrics that are numerically simulated by probing the model., Comment: https://github.com/ENSTA-U2IS-AI/Dataset_XAI

Random forests utilize bootstrap sampling to create an individual training set for each component tree. This involves sampling with replacement, with the number of instances equal to the size of the original training set ($N$). Research literature indicates that drawing fewer than $N$ observations can also yield satisfactory results. The ratio of the number of observations in each bootstrap sample to the total number of training instances is called the bootstrap rate (BR). Sampling more than $N$ observations (BR $>$ 1) has been explored in the literature only to a limited extent and has generally proven ineffective. In this paper, we re-examine this approach using 36 diverse datasets and consider BR values ranging from 1.2 to 5.0. Contrary to previous findings, we show that such parameterization can result in statistically significant improvements in classification accuracy compared to standard settings (BR $\leq$ 1). Furthermore, we investigate what the optimal BR depends on and conclude that it is more a property of the dataset than a dependence on the random forest hyperparameters. Finally, we develop a binary classifier to predict whether the optimal BR is $\leq$ 1 or $>$ 1 for a given dataset, achieving between 81.88\% and 88.81\% accuracy, depending on the experiment configuration.

The tunability and spatial precision of paramagnetic molecules makes them attractive for quantum sensing. However, usual microwave-based detection methods have poor temporal and spatial resolution, and optical methods compatible with room-temperature solutions have remained elusive. Here, we utilize pump-probe polarization spectroscopy to initialize and track electron spin coherence in a molecule. Designed to efficiently couple spins to light, aqueous $K_{2}IrCl_{6}$ enables detection of few-picosecond free induction decay at room temperature and micromolar concentrations. Viscosity is found to strongly vary decoherence lifetimes. This work redefines the meaning of room-temperature coherence by improving experimental time resolution by up to five orders of magnitude. Doing so unveils a new regime of electron spin coherence, opening the door to new synthetic design and applications of molecular quantum bits.

High-spin molecules allow for bottom-up qubit design and are promising platforms for magnetic sensing and quantum information science. Optical addressability of molecular electron spins has also been proposed in first-row transition metal complexes via optically-detected magnetic resonance (ODMR) mechanisms analogous to the diamond-NV colour centre. However, significantly less progress has been made on the front of metal-free molecules, which can deliver lower costs and milder environmental impacts. At present, most luminescent open-shell organic molecules are $\pi$-diradicals, but such systems often suffer from poor ground-state open-shell characters necessary to realise a stable ground-state molecular qubit. In this work, we use alternancy symmetry to selectively minimise radical-radical interactions in the ground state, generating $\pi$-systems with high diradical characters. We call them m-dimers, referencing the need to covalently link two benzylic radicals at their meta carbon atoms for the desired symmetry. Through a detailed electronic structure analysis, we find that the excited states of alternant hydrocarbon m-diradicals contain important symmetries that can be used to construct ODMR mechanisms leading to ground-state spin polarisation. The molecular parameters are set in the context of a tris(2,4,6-trichlorophenyl)methyl (TTM) radical dimer covalently tethered at the meta position, demonstrating the feasibility of alternant m-diradicals as molecular colour centres., Comment: 12 pages, 5 figures, 1 table. Minor edits made to the manuscript text

In this paper, we introduce an explainable algorithm designed from a multi-modal foundation model, that performs fast and explainable image classification. Drawing inspiration from CLIP-based Concept Bottleneck Models (CBMs), our method creates a latent space where each neuron is linked to a specific word. Observing that this latent space can be modeled with simple distributions, we use a Mixture of Gaussians (MoG) formalism to enhance the interpretability of this latent space. Then, we introduce CLIP-QDA, a classifier that only uses statistical values to infer labels from the concepts. In addition, this formalism allows for both local and global explanations. These explanations come from the inner design of our architecture, our work is part of a new family of greybox models, combining performances of opaque foundation models and the interpretability of transparent models. Our empirical findings show that in instances where the MoG assumption holds, CLIP-QDA achieves similar accuracy with state-of-the-art methods CBMs. Our explanations compete with existing XAI methods while being faster to compute.

Context. Multi-wavelength studies of galaxies and galactic nuclei allow us to build a relatively more complete picture of the interstellar medium (ISM), especially in the dusty regions of starburst galaxies. An understanding of the physical processes in nearby galaxies can assist in the study of more distant sources at higher redshifts, which cannot be resolved. Aims. We aimed to use observations presented in the first part of this series of papers to model the physical conditions of the ISM in the nuclear region of NGC 253, in order to obtain primary parameters such as gas densities and metallicities. From the created model we further calculated secondary parameters such as gas masses of the different phases, and estimated the fraction of [C II] 158 um from the different phases, which allowed us to probe the nuclear star-formation rate. Methods. To compare theory with our observations we used MULTIGRIS, a probabilistic tool that determines probabilities for certain ISM parameters from a grid of Cloudy models together with a set of spectroscopic lines. Results. We find that the hypothetical active galactic nucleus within NGC 253 has only a minor impact compared to the starburst on the heating of the ISM as probed by the observed lines. We characterise the ISM and obtain parameters such as a solar metallicity, a mean density of ~230cm-3 , an ionisation parameter of log U = -3, and an age of the nuclear cluster of ~2 Myr. Furthermore, we estimate the masses of the ionised (3.8 x 10^6 M_sol ), neutral atomic (9.1 x 10^6 M_sol ), and molecular (2.0 x 10^8 M_sol ) gas phases as well as the dust mass (1.8 x 10^6 M_sol ) in the nucleus of NGC 253., Comment: Accepted for publication in A&A; 15 pages, 10 Figures, 6 Tables

Graphene flow sensors hold great prospects for applications, but also encounter many difficulties, such as unwanted electrochemical phenomena, low measurable signal and limited dependence on the flow direction. This study proposes a novel approach allowing for the detection of a flow direction-dependent electric signal in aqueous solutions of salts, acids and bases. The key element in the proposed solution is the use of a reference electrode which allows external gating of the graphene structure. Using external gating enables to enhance substantially the amplitude of the flow-generated signal. Simultaneous measurement of the reference electrode current allows us to recover a flow-direction-sensitive component of the flow-induced voltage in graphene. The obtained results are discussed in terms of the Coulomb interaction and other phenomena which can be present at the interface of graphene with the aqueous solution., Comment: 7 pages, 6 figures

Semantic segmentation methods have advanced significantly. Still, their robustness to real-world perturbations and object types not seen during training remains a challenge, particularly in safety-critical applications. We propose a novel approach to improve the robustness of semantic segmentation techniques by leveraging the synergy between label-to-image generators and image-to-label segmentation models. Specifically, we design Robusta, a novel robust conditional generative adversarial network to generate realistic and plausible perturbed images that can be used to train reliable segmentation models. We conduct in-depth studies of the proposed generative model, assess the performance and robustness of the downstream segmentation network, and demonstrate that our approach can significantly enhance the robustness in the face of real-world perturbations, distribution shifts, and out-of-distribution samples. Our results suggest that this approach could be valuable in safety-critical applications, where the reliability of perception modules such as semantic segmentation is of utmost importance and comes with a limited computational budget in inference. We release our code at https://github.com/ENSTA-U2IS-AI/robusta., Comment: Published as a conference paper at WACV 2024

(abridged) Far-infrared (FIR) line emission provides key information about the gas cooling and heating due to shocks and UV radiation associated with the early stages of star formation. Gas cooling via FIR lines might, however, depend on metallicity. We aim to quantify the FIR line emission and determine the spatial distribution of the CO rotational temperature, ultraviolet (UV) radiation field, and H2 number density toward the embedded cluster Gy 3-7 in the CMa-l224 star-forming region, whose metallicity is expected to be intermediate between that of the LMC and the Solar neighborhood. By comparing the total luminosities of CO and [O I] toward Gy 3-7 with values found for low- and high-mass protostars extending over a broad range of metallicities, we also aim to identify the possible effects of metallicity on the FIR line cooling within our Galaxy. We studied SOFIA/FIFI-LS spectra of Gy 3-7 covering several FIR lines. The spatial extent of CO high-J (J>14) emission resembles that of the elongated 160 um continuum emission detected with Herschel. The CO transitions from J=14-13 to J=16-15 are detected throughout the cluster and show a median rotational temperature of 170+/-30 K on Boltzmann diagrams. Comparisons to other protostars observed with Herschel show a good agreement with intermediate-mass sources in the inner Galaxy. Assuming an origin of the [O I] and high-J CO emission in UV-irradiated C-shocks, we obtained pre-shock H2 number densities of 10^4-5 cm-3 and UV radiation field strengths of 0.1-10 Habing fields. Far-IR line observations reveal ongoing star formation in Gy 3-7, dominated by intermediate-mass Class 0/I young stellar objects. The ratio of molecular-to-atomic far-IR line emission shows a decreasing trend with bolometric luminosities of the protostars. However, it does not indicate that the low-metallicity has an impact on the line cooling in Gy 3-7., Comment: 26 pages, 23 figures

Context. Massive star-formation leads to enrichment with heavy elements of the interstellar medium. On the other hand, the abundance of heavy elements is a key parameter to study the star-formation history of galaxies. Furthermore, the total molecular hydrogen mass, usually determined by converting CO or [C ii] 158 $\mu$m luminosities, depends on the metallicity as well. The excitation of metallicity-sensitive emission lines, however, depends on the gas density of H ii regions, where they arise. Aims. We used spectroscopic observations from SOFIA, Herschel, and Spitzer of the nuclear region of the starburst galaxy NGC 253, as well as photometric observations from GALEX, 2MASS, Spitzer, and Herschel in order to derive physical properties such as the optical depth to correct for extinction, as well as the gas density and metallicity of the central region. Methods. Ratios of the integrated line fluxes of several species were utilised to derive the gas density and metallicity. The [O iii] along with the [S iii] and [N ii] line flux ratios for example, are sensitive to the gas density but nearly independent of the local temperature. As these line ratios trace different gas densities and ionisation states, we examined if these lines may originate from different regions within the observing beam. The ([Ne ii] 13 $\mu$m + [Ne iii] 16 $\mu$m)/Hu $\alpha$ line flux ratio on the other hand, is independent of the depletion onto dust grains but sensitive to the Ne/H abundance ratio and will be used as a tracer for metallicity of the gas. Results. We derived values for gas phase abundances of the most important species, as well as estimates for the optical depth and the gas density of the ionised gas in the nuclear region of NGC 253. We obtained densities of at least two different ionised components $(< 84$ cm$^{-3}$ and $\sim 170 - 212$ cm$^{-3})$ and a metallicity of solar value., Comment: 12 pages, 10 Figures, 6 Tables. Accepted for publication in A&A

Two-dimensional (2D) magnetic crystals hold promise for miniaturized and ultralow power electronic devices that exploit spin manipulation. In these materials, large, controllable magnetocrystalline anisotropy is a prerequisite for the stabilization and manipulation of long-range magnetic order. In known 2D magnetic crystals, relatively weak magnetocrystalline anisotropy results in typically soft ferromagnetism. Here, we demonstrate that ferromagnetic order persists down to the thinnest limit of Fe$_x$TaS$_2$ (Fe-intercalated bilayer 2H-TaS$_2$) with giant coercivities up to 3 tesla. We prepare Fe-intercalated TaS$_2$ by chemical intercalation of van der Waals layered 2H-TaS$_2$ crystals and perform variable-temperature quantum transport, transmission electron microscopy, and confocal Raman spectroscopy measurements to shed new light on the coupled effects of dimensionality, degree of intercalation, and intercalant order/disorder on the hard ferromagnetic behavior of Fe$_x$TaS$_2$. More generally, we show that chemical intercalation gives access to a rich synthetic parameter space for low-dimensional magnets, in which magnetic properties can be tailored by the choice of the host material and intercalant identity/amount, in addition to the manifold distinctive degrees of freedom available in atomically thin, van der Waals crystals.

Predictive uncertainty estimation is essential for safe deployment of Deep Neural Networks in real-world autonomous systems. However, disentangling the different types and sources of uncertainty is non trivial for most datasets, especially since there is no ground truth for uncertainty. In addition, while adverse weather conditions of varying intensities can disrupt neural network predictions, they are usually under-represented in both training and test sets in public datasets.We attempt to mitigate these setbacks and introduce the MUAD dataset (Multiple Uncertainties for Autonomous Driving), consisting of 10,413 realistic synthetic images with diverse adverse weather conditions (night, fog, rain, snow), out-of-distribution objects, and annotations for semantic segmentation, depth estimation, object, and instance detection. MUAD allows to better assess the impact of different sources of uncertainty on model performance. We conduct a thorough experimental study of this impact on several baseline Deep Neural Networks across multiple tasks, and release our dataset to allow researchers to benchmark their algorithm methodically in adverse conditions. More visualizations and the download link for MUAD are available at https://muad-dataset.github.io/., Comment: Accepted at BMVC 2022

Several metrics exist to quantify the similarity between images, but they are inefficient when it comes to measure the similarity of highly distorted images. In this work, we propose to empirically investigate perceptual metrics based on deep neural networks for tackling the Image Quality Assessment (IQA) task. We study deep perceptual metrics according to different hyperparameters like the network's architecture or training procedure. Finally, we propose our multi-resolution perceptual metric (MR-Perceptual), that allows us to aggregate perceptual information at different resolutions and outperforms standard perceptual metrics on IQA tasks with varying image deformations. Our code is available at https://github.com/ENSTA-U2IS/MR_perceptual

Observations that require large physical instrument dimensions and/or a considerable amount of cryogens, as it is the case for high spatial resolution far infrared (FIR) astronomy, currently still face technological limits for their execution from space. Angular resolution and available observational capabilities are particularly affected. Balloon-based platforms promise to complement the existing observational capabilities by offering means to deploy comparatively large telescopes with comparatively little effort, including other advantages such as the possibility to regularly refill cryogens and to change and/or update instruments. The planned European Stratospheric Balloon Observatory (ESBO) aims at providing these additional large aperture FIR capabilities, exceeding the spatial resolution of Herschel, in the long term. The plans focus on reusable platforms performing regular flights and an operations concept that provides researchers with proposal-based access to observations. It thereby aims at offering a complement to other airborne, ground-based and space-based observatories in terms of access to wavelength regions, spatial resolution capability, and photometric stability. While the FIR capabilities are a main long-term objective, ESBO will offer benefits in other wavelength regimes along the way. Within the ESBO Design Study (ESBO DS), a prototype platform carrying a 0.5 m telescope for ultraviolet and visible light observations is being built and a platform concept for a next-generation FIR telescope is being studied. A flight of the UV/VIS prototype platform is estimated for 2021. In this paper we will outline the scientific and technical motivation for a large aperture balloon-based FIR observatory and the ESBO DS approach towards such an infrastructure. Secondly, we will present the technical motivation, science case, and instrumentation of the 0.5 m UV/VIS platform., Comment: Proc. 69th International Astronautical Congress, 1-5 October 2018. arXiv admin note: text overlap with arXiv:2111.11068

Facial dysmorphology or malocclusion is frequently associated with abnormal growth of the face. The ability to predict facial growth (FG) direction would allow clinicians to prepare individualized therapy to increase the chance for successful treatment. Prediction of FG direction is a novel problem in the machine learning (ML) domain. In this paper, we perform feature selection and point the attribute that plays a central role in the abovementioned problem. Then we successfully apply data augmentation (DA) methods and improve the previously reported classification accuracy by 2.81%. Finally, we present the results of two experienced clinicians that were asked to solve a similar task to ours and show how tough is solving this problem for human experts.

First attempts of prediction of the facial growth (FG) direction were made over half of a century ago. Despite numerous attempts and elapsed time, a satisfactory method has not been established yet and the problem still poses a challenge for medical experts. To our knowledge, this paper is the first Machine Learning approach to the prediction of FG direction. Conducted data analysis reveals the inherent complexity of the problem and explains the reasons of difficulty in FG direction prediction based on 2D X-ray images. To perform growth forecasting, we employ a wide range of algorithms, from logistic regression, through tree ensembles to neural networks and consider three, slightly different, problem formulations. The resulting classification accuracy varies between 71% and 75%.

Van der Waals heteroepitaxy allows deterministic control over lattice mismatch or azimuthal orientation between atomic layers to produce long wavelength superlattices. The resulting electronic phases depend critically on the superlattice periodicity as well as localized structural deformations that introduce disorder and strain. Here, we introduce Bragg interferometry, based on four-dimensional scanning transmission electron microscopy, to capture atomic displacement fields in twisted bilayer graphene with twist angles < 2{\deg}. Nanoscale spatial fluctuations in twist angle and uniaxial heterostrain are statistically evaluated, revealing the prevalence of short-range disorder in this class of materials. By quantitatively mapping strain tensor fields we uncover two distinct regimes of structural relaxation -- in contrast to previous models depicting a single continuous process -- and we disentangle the electronic contributions of the rotation modes that comprise this relaxation. Further, we find that applied heterostrain accumulates anisotropically in saddle point regions to generate distinctive striped shear strain phases. Our results thus establish the reconstruction mechanics underpinning the twist angle dependent electronic behaviour of twisted bilayer graphene, and provide a new framework for directly visualizing structural relaxation, disorder, and strain in any moir\'e material., Comment: 29 pages, 6 figures plus supporting information (42 pages, 28 figures)

Image classification has become a ubiquitous task. Models trained on good quality data achieve accuracy which in some application domains is already above human-level performance. Unfortunately, real-world data are quite often degenerated by the noise existing in features and/or labels. There are quite many papers that handle the problem of either feature or label noise, separately. However, to the best of our knowledge, this piece of research is the first attempt to address the problem of concurrent occurrence of both types of noise. Basing on the MNIST, CIFAR-10 and CIFAR-100 datasets, we experimentally proved that the difference by which committees beat single models increases along with noise level, no matter it is an attribute or label disruption. Thus, it makes ensembles legitimate to be applied to noisy images with noisy labels. The aforementioned committees' advantage over single models is positively correlated with dataset difficulty level as well. We propose three committee selection algorithms that outperform a strong baseline algorithm which relies on an ensemble of individual (nonassociated) best models.

We present detailed Raman studies of graphene deposited on gallium nitride nanowires with different variations in height. Our results show that different density and height of nanowires being in contact with graphene impact graphene properties like roughness, strain and carrier concentration as well as density and type of induced defects. Detailed analysis of Raman spectra of graphene deposited on different nanowire substrates shows that bigger differences in nanowires height increase graphene strain, while higher number of nanowires in contact with graphene locally reduce the strain. Moreover, the value of graphene carrier concentration is found to be correlated with the density of nanowires in contact with graphene. Analysis of intensity ratios of Raman G, D and D' bands enable to trace how nanowire substrate impacts the defect concentration and type. The lowest concentration of defects is observed for graphene deposited on nanowires of the lowest density. Contact between graphene and densely arranged nanowires leads to a large density of vacancies. On the other hand, grain boundaries are the main type of defects in graphene on rarely distributed nanowires. Our results also show modification of graphene carrier concentration and strain by different types of defects present in graphene.

A constant height of gallium nitride (GaN) nanowires with graphene deposited on them is shown to have a strong enhancement of Raman scattering, whilst variable height nanowires fail to give such an enhancement. Scanning electron microscopy reveals a smooth graphene surface which is present when the GaN nanowires are uniform, whereas graphene on nanowires with substantial height differences is observed to be pierced and stretched by the uppermost nanowires. The energy shifts of the characteristic Raman bands confirms that these differences in the nanowire height has a significant impact on the local graphene strain and the carrier concentration. The images obtained by Kelvin probe force microscopy show clearly that the carrier concentration in graphene is modulated by the nanowire substrate and dependent on the nanowire density. Therefore, the observed surface enhanced Raman scattering for graphene deposited on GaN nanowires of comparable height is triggered by self-induced nano-gating to the graphene. However, no clear correlation of the enhancement with the strain or the carrier concentration of graphene was discovered.