Search

Under high perpendicular magnetic field and at low temperatures, graphene develops an insulating state at the charge neutrality point. This state, dubbed $\nu=0$, is due to the interplay between electronic interactions and the four-fold spin and valley degeneracies in the flat band formed by the $n=0$ Landau level. Determining the ground state of $\nu=0$, including its spin and valley polarization, has been a theoretical and experimental undertaking for almost two decades. Here, we present experiments probing the bulk thermal transport properties of monolayer graphene at $\nu=0$, which directly probe its ground state and collective excitations. We observe a vanishing bulk thermal transport, in contradiction with the expected ground state, predicted to have a finite thermal conductance even at very low temperature. Our result highlight the need for further investigations on the nature of $\nu=0$., Comment: Contains Supplementary Information

Graph Neural Networks (GNNs) have become increasingly popular for effectively modeling data with graph structures. Recently, attention mechanisms have been integrated into GNNs to improve their ability to capture complex patterns. This paper presents the first comprehensive study revealing a critical, unexplored consequence of this integration: the emergence of Massive Activations (MAs) within attention layers. We introduce a novel method for detecting and analyzing MAs, focusing on edge features in different graph transformer architectures. Our study assesses various GNN models using benchmark datasets, including ZINC, TOX21, and PROTEINS. Key contributions include (1) establishing the direct link between attention mechanisms and MAs generation in GNNs, (2) developing a robust definition and detection method for MAs based on activation ratio distributions, (3) introducing the Explicit Bias Term (EBT) as a potential countermeasure and exploring it as an adversarial framework to assess models robustness based on the presence or absence of MAs. Our findings highlight the prevalence and impact of attention-induced MAs across different architectures, such as GraphTransformer, GraphiT, and SAN. The study reveals the complex interplay between attention mechanisms, model architecture, dataset characteristics, and MAs emergence, providing crucial insights for developing more robust and reliable graph models.

The junction temperature is a very important parameter for monitoring power electronics converters based on MOSFET transistors. They offer the possibility of switching at relatively higher frequencies than other transistors like IGTBTs. However, the electrical parameters of MOSFETs are highly thermally dependent. The thermo-dependence of MOSFET electrical parameters is rarely taken into consideration when implementing control strategies, for many technological reasons, such as the difficulty of measuring the junction temperature. In practice, the junction temperature of transistors is inaccessible for direct measurement. The presence of a gel covering the chips, that provides electrical and thermal insulation, makes measurement by infrared thermography impossible. Furthermore, direct thermocouple measurement cannot be implemented due to the electromagnetic disturbances in the environment. Several researchers have attempted to correlate chip temperature with thermosensitive electrical parameters. In the present work, a thermal convolutive model has been developed to estimate the junction temperatures of two MOSFET transistors belonging to the same electronic circuit from external temperature measurements in two well-chosen locations (far away enough from the junction to avoid electromagnetic interference), using also the measured power dissipated on each chip. The thermal coupling between the two transistors has been considered in the form of mutual transmittances. The model was first calibrated using three-dimensional numerical simulations in COMSOL Multiphysics, followed by an experimental study. The results are very promising, illustrating the robustness of the convolutional model.

In the complex landscape of hematologic samples such as peripheral blood or bone marrow derived from flow cytometry (FC) data, cell-level prediction presents profound challenges. This work explores injecting hierarchical prior knowledge into graph neural networks (GNNs) for single-cell multi-class classification of tabular cellular data. By representing the data as graphs and encoding hierarchical relationships between classes, we propose our hierarchical plug-in method to be applied to several GNN models, namely, FCHC-GNN, and effectively designed to capture neighborhood information crucial for single-cell FC domain. Extensive experiments on our cohort of 19 distinct patients, demonstrate that incorporating hierarchical biological constraints boosts performance significantly across multiple metrics compared to baseline GNNs without such priors. The proposed approach highlights the importance of structured inductive biases for gaining improved generalization in complex biological prediction tasks., Comment: 14 pages, ICML Conference Workshop 2024. arXiv admin note: text overlap with arXiv:2402.18610

In this paper we consider stochastic Fokker-Planck Partial Differential Equations (PDEs), obtained as the mean-field limit of weakly interacting particle systems subjected to both independent (or idiosyncratic) and common Brownian noises. We provide sufficient conditions under which the deterministic counterpart of the Fokker-Planck equation, which corresponds to particle systems that are just subjected to independent noises, has several invariant measures, but for which the stochastic version admits a unique invariant measure under the presence of the additive common noise. The very difficulty comes from the fact that the common noise is just of finite dimension while the state variable, which should be seen as the conditional marginal law of the system given the common noise, lives in a space of infinite dimension. In this context, our result holds true if, in addition to standard confining properties, the mean field interaction term forces the system to be attracted by its conditional mean given the common noise and the intensity of the idiosyncratic noise is small.

We introduce a new approach for estimating the invariant density of a multidimensional diffusion when dealing with high-frequency observations blurred by independent noises. We consider the intermediate regime, where observations occur at discrete time instances $k\Delta_n$ for $k=0,\dots,n$, under the conditions $\Delta_n\to 0$ and $n\Delta_n\to\infty$. Our methodology involves the construction of a kernel density estimator that uses a pre-averaging technique to proficiently remove noise from the data while preserving the analytical characteristics of the underlying signal and its asymptotic properties. The rate of convergence of our estimator depends on both the anisotropic regularity of the density and the intensity of the noise. We establish conditions on the intensity of the noise that ensure the recovery of convergence rates similar to those achievable without any noise. Furthermore, we prove a Bernstein concentration inequality for our estimator, from which we derive an adaptive procedure for the kernel bandwidth selection.

Atom probe tomography data is composed of a list of coordinates of the reconstructed atoms in the probed volume. The elemental identity of each atom is derived from time-of-flight mass spectrometry, with no local energetic or chemical information readily available within the mass spectrum. Here, we used a new data processing technique referred to as field evaporation energy loss spectroscopy (FEELS), which analyses the tails of mass peaks. FEELS was used to extract critical energetic parameters that characterize the field evaporation process, which are related to the binding energy of atoms to the surface under intense electrostatic field and dependent of the path followed by the departing atoms during the field evaporation process. We focused our study on different pure face centered cubic metals (Al, Ni, Rh). We demonstrate that the energetic parameters extracted from mass spectra can be mapped in 2D with nanometric resolution. A dependence on the considered crystallographic planes is observed, with sets of planes of low Miller indices showing a lower sensitivity to the intensity of the electric field, which indicates a lower effective attachment energy. The temperature is also an important parameter in particular for Al, which we attribute to an energetic transition between two paths of field evaporation between 25K and 60K close to (002) pole at the specimen's surface. This paper shows that the complex information that can be retrieved from the measured energy loss of surface atoms is important both instrumentally and fundamentally.

This paper presents FlowCyt, the first comprehensive benchmark for multi-class single-cell classification in flow cytometry data. The dataset comprises bone marrow samples from 30 patients, with each cell characterized by twelve markers. Ground truth labels identify five hematological cell types: T lymphocytes, B lymphocytes, Monocytes, Mast cells, and Hematopoietic Stem/Progenitor Cells (HSPCs). Experiments utilize supervised inductive learning and semi-supervised transductive learning on up to 1 million cells per patient. Baseline methods include Gaussian Mixture Models, XGBoost, Random Forests, Deep Neural Networks, and Graph Neural Networks (GNNs). GNNs demonstrate superior performance by exploiting spatial relationships in graph-encoded data. The benchmark allows standardized evaluation of clinically relevant classification tasks, along with exploratory analyses to gain insights into hematological cell phenotypes. This represents the first public flow cytometry benchmark with a richly annotated, heterogeneous dataset. It will empower the development and rigorous assessment of novel methodologies for single-cell analysis., Comment: arXiv admin note: text overlap with arXiv:2402.18611

In the complex landscape of hematologic samples such as peripheral blood or bone marrow, cell classification, delineating diverse populations into a hierarchical structure, presents profound challenges. This study presents LeukoGraph, a recently developed framework designed explicitly for this purpose employing graph attention networks (GATs) to navigate hierarchical classification (HC) complexities. Notably, LeukoGraph stands as a pioneering effort, marking the application of graph neural networks (GNNs) for hierarchical inference on graphs, accommodating up to one million nodes and millions of edges, all derived from flow cytometry data. LeukoGraph intricately addresses a classification paradigm where for example four different cell populations undergo flat categorization, while a fifth diverges into two distinct child branches, exemplifying the nuanced hierarchical structure inherent in complex datasets. The technique is more general than this example. A hallmark achievement of LeukoGraph is its F-score of 98%, significantly outclassing prevailing state-of-the-art methodologies. Crucially, LeukoGraph's prowess extends beyond theoretical innovation, showcasing remarkable precision in predicting both flat and hierarchical cell types across flow cytometry datasets from 30 distinct patients. This precision is further underscored by LeukoGraph's ability to maintain a correct label ratio, despite the inherent challenges posed by hierarchical classifications.

Despite the promising results of machine learning models in malicious files detection, they face the problem of concept drift due to their constant evolution. This leads to declining performance over time, as the data distribution of the new files differs from the training one, requiring frequent model update. In this work, we propose a model-agnostic protocol to improve a baseline neural network against drift. We show the importance of feature reduction and training with the most recent validation set possible, and propose a loss function named Drift-Resilient Binary Cross-Entropy, an improvement to the classical Binary Cross-Entropy more effective against drift. We train our model on the EMBER dataset, published in2018, and evaluate it on a dataset of recent malicious files, collected between 2020 and 2023. Our improved model shows promising results, detecting 15.2% more malware than a baseline model.

The increased time- and length-scale of classical molecular dynamics simulations have led to raw data flows surpassing storage capacities, necessitating on-the-fly integration of structural analysis algorithms. As a result, algorithms must be computationally efficient, accurate, and stable at finite temperature to reliably extract the relevant features of the data at simulation time. In this work, we leverage spectral descriptors to encode local atomic environments and build crystal structure classification models. In addition to the classical way spectral descriptors are computed, i.e. over a fixed radius neighborhood sphere around a central atom, we propose an extension to make them independent from the material's density. Models are trained on defect-free crystal structures with moderate thermal noise and elastic deformation, using the linear discriminant analysis (LDA) method for dimensionality reduction and logistic regression (LR) for subsequent classification. The proposed classification model is intentionally designed to be simple, incorporating only a limited number of parameters. This deliberate simplicity enables the model to be trained effectively even when working with small databases. Despite the limited training data, the model still demonstrates inherent transferability, making it applicable to a broader range of scenarios and datasets. The accuracy of our models in extreme conditions is compared to traditional algorithms from the literature, namely adaptive common neighbor analysis (a-CNA), polyhedral template matching (PTM) and diamond structure identification (IDS). Finally, we showcase two applications of our method: tracking a solid-solid BCC-to-HCP phase transformation in Zirconium at high pressure up to high temperature, and visualizing stress-induced dislocation loop expansion in single crystal FCC Aluminum containing a Frank-Read source, at high temperature.

Accurate structural analysis is essential to gain physical knowledge and understanding of atomic-scale processes in materials from atomistic simulations. However, traditional analysis methods often reach their limits when applied to crystalline systems with thermal fluctuations, defect-induced distortions, partial vitrification, etc. In order to enhance the means of structural analysis, we present a novel descriptor for encoding atomic environments into 2D images, based on a pixelated representation of graph-like architecture with weighted edge connections of neighboring atoms. This descriptor is well adapted for Convolutional Neural Networks and enables accurate structural analysis at a low computational cost. In this paper, we showcase a series of applications, including the classification of crystalline structures in distorted systems, tracking phase transformations up to the melting temperature, and analyzing liquid-to-amorphous transitions in pure metals and alloys. This work provides the foundation for robust and efficient structural analysis in materials science, opening up new possibilities for studying complex structural processes, which can not be described with traditional approaches.

This paper presents an investigation into the long-term behaviour of solutions to a nonlinear stochastic McKean-Vlasov equation with common noise. The equation arises naturally in themean-field limit of systems composed of interacting particles subject to both idiosyncratic and common noise. Initially, we demonstrate that the addition of common diffusion in each particle's dynamics does not disrupt the established stability results observed in the absence of common noise.However, our main objective is to understand how the presence of common noise can restore theuniqueness of equilibria. Specifically, in a non-convex landscape, we establish uniqueness and convergence towards equilibria under two specific conditions: (1) when the dimension of the ambientspace equals 1, and (2) in the absence of idiosyncratic noise in the system.

Deep hashing has shown to be a complexity-efficient solution for the Approximate Nearest Neighbor search problem in high dimensional space. Many methods usually build the loss function from pairwise or triplet data points to capture the local similarity structure. Other existing methods construct the similarity graph and consider all points simultaneously. Auto-encoding Twin-bottleneck Hashing is one such method that dynamically builds the graph. Specifically, each input data is encoded into a binary code and a continuous variable, or the so-called twin bottlenecks. The similarity graph is then computed from these binary codes, which get updated consistently during the training. In this work, we generalize the original model into a supervised deep hashing network by incorporating the label information. In addition, we examine the differences of codes structure between these two networks and consider the class imbalance problem especially in multi-labeled datasets. Experiments on three datasets yield statistically significant improvement against the original model. Results are also comparable and competitive to other supervised methods.

We report on the theoretical derivation of macroscopic thermal properties (specific heat, thermal conductivity) of an electrically insulating rod connected to two reservoirs, from the linear superposition of its mechanical mode Brownian motions. The calculation is performed for a weak thermal gradient, in the classical limit (high temperature). The development is kept basic as far as geometry and experimental conditions are concerned, enabling an almost fully analytic treatment. In the modeling, each of the modes is subject to a specific Langevin force, which enables to produce the required temperature profile along the rod. The theory is predictive: the temperature gradient (and therefore energy transport) is linked to motion amplitude cross-correlations between nearby mechanical modes. This arises because energy transport is actually mediated by mixing between the modal waves, and not by the modes themselves. This result can be tested on experiments, and shall extend the concepts underlying equipartition and fluctuation-dissipation theorems. The theory links intimately the macroscopic size of the clamping region where the mixing occurs to the microscopic lengthscale of the problem at hand: the phonon mean-free-path. This clamping region, which is key, has received recently a renewed attention in the field of nanomechanics with topical works on "phonon shields" and "soft clamping". We believe that our work should impact the domain of thermal transport in nanostructures, with future developments of the theory toward the quantum regime.

The Josephson junction is a building block of quantum circuits. Its behavior, well understood when treated as an isolated entity, is strongly affected by coupling to an electromagnetic environment. In 1983, Schmid predicted that a Josephson junction shunted by a resistance exceeding the resistance quantum $\mathbf{\textit{R}}_\mathrm{Q} = h/4e^2 \approx 6.45$ k$\mathbf{\Omega}$ for Cooper pairs would become insulating since the phase fluctuations would destroy the coherent Josephson coupling. However, recent microwave measurements have questioned this interpretation. Here, we insert a small Josephson junction in a Johnson-Nyquist-type setup where it is driven by weak current noise arising from thermal fluctuations. Our heat probe minimally perturbs the junction's equilibrium, shedding light on features not visible in charge transport. We find that the Josephson critical current completely vanishes in DC charge transport measurement, and the junction demonstrates Coulomb blockade in agreement with the theory. Surprisingly, thermal transport measurements show that the Josephson junction acts as an inductor at high frequencies, unambiguously demonstrating that a supercurrent survives despite the Coulomb blockade observed in DC measurements. The discrepancy between these two measurements highlights the difference between the low and the high frequency response of a junction and calls for further theoretical and experimental inputs on the dynamics of Josephson junctions \textcolor{black}{operating at high frequencies in highly resistive environments., Comment: Final version accepted for publication

Recent pioneering works have set the stage for exploring anyon braiding statistics from negative current cross-correlations along two intersecting quasiparticle beams. In such a dual-source - analyzer quantum point contact setup, also referred to as 'collider', the anyon exchange phase of fractional quantum Hall quasiparticles is predicted to be imprinted into the cross-correlations characterized by an effective Fano-factor $P$. In the case of symmetric incoming quasiparticle beams, conventional fermions result in a vanishing $P$. In marked contrast, we observe signatures of anyon statistics in the negative $P$ found both for the $e/3$ Laughlin quasiparticles at filling factor $\nu=1/3$ ($P\approx-2$, corroborating previous findings), and for the $e/5$ quasiparticles in the hierarchical state $\nu=2/5$ ($P\approx-1$). Nevertheless, we argue that the quantitative connection between numerical value of $P\neq0$ and specific fractional exchange phase is hampered by the influence of the analyzer conductance dependence on the voltages used to generate the quasiparticles. Finally, we address the important challenge how to distinguish at $\nu=1/3$ between negative cross-correlations induced by a fractional braid phase, and those resulting from a different Andreev-like mechanism. Although with symmetric sources $P$ does not exhibit signatures of a crossover when the analyzer is progressively detuned to favor Andreev processes, we demonstrate that changing the balance between sources provides a mean to discriminate between the two mechanisms.

Crop diseases significantly affect the quantity and quality of agricultural production. In a context where the goal of precision agriculture is to minimize or even avoid the use of pesticides, weather and remote sensing data with deep learning can play a pivotal role in detecting crop diseases, allowing localized treatment of crops. However, combining heterogeneous data such as weather and images remains a hot topic and challenging task. Recent developments in transformer architectures have shown the possibility of fusion of data from different domains, for instance text-image. The current trend is to custom only one transformer to create a multimodal fusion model. Conversely, we propose a new approach to realize data fusion using three transformers. In this paper, we first solved the missing satellite images problem, by interpolating them with a ConvLSTM model. Then, proposed a multimodal fusion architecture that jointly learns to process visual and weather information. The architecture is built from three main components, a Vision Transformer and two transformer-encoders, allowing to fuse both image and weather modalities. The results of the proposed method are promising achieving 97\% overall accuracy.