The present work aims at exploring the scale-by-scale kinetic energy exchanges in multiphase turbulence. For this purpose, we derive the Kármán-Howarth-Monin equation which accounts for the variations of density and viscosity across the two phases together with the effect of surface tension. We consider both conventional and phase conditional averaging operators. This framework is applied to numerical data from detailed simulations of forced homogeneous and isotropic turbulence covering different values for the liquid volume fraction, the liquid/gas density ratio, the Reynolds, and Weber numbers. We confirm the existence of an additional transfer term due to surface tension. Part of the kinetic energy injected at large scales is transferred into kinetic energy at smaller scales by classical non-linear transport while another part is transferred to surface energy before being released back into kinetic energy, but at smaller scales. The overall kinetic energy transfer rate is larger than in single phase flows. Kinetic energy budgets conditioned in a given phase show that the scale-by-scale transport of turbulent kinetic energy due to pressure is a gain (loss) of kinetic energy for the lighter (heavier) phase. Its contribution can be dominant when the gas volume fraction becomes small or when the density ratio increases. Building on previous work, we hypothesize the existence of a pivotal scale above which kinetic energy is stored into surface deformation and below which the kinetic energy is released by interface restoration. Some phenomenological predictions for this scale are discussed.

We present COMEDIAN, a novel pipeline to initialize spatiotemporal transformers for action spotting, which involves self-supervised learning and knowledge distillation. Action spotting is a timestamp-level temporal action detection task. Our pipeline consists of three steps, with two initialization stages. First, we perform self-supervised initialization of a spatial transformer using short videos as input. Additionally, we initialize a temporal transformer that enhances the spatial transformer's outputs with global context through knowledge distillation from a pre-computed feature bank aligned with each short video segment. In the final step, we fine-tune the transformers to the action spotting task. The experiments, conducted on the SoccerNet-v2 dataset, demonstrate state-of-the-art performance and validate the effectiveness of COMEDIAN's pretraining paradigm. Our results highlight several advantages of our pretraining pipeline, including improved performance and faster convergence compared to non-pretrained models.

We consider linear differential-algebraic equations DAEs and the Kronecker canonical form KCF of the corresponding matrix pencils. We also consider linear control systems and their Morse canonical form MCF. For a linear DAE, a procedure named explicitation is proposed, which attaches to any linear DAE a linear control system defined up to a coordinates change, a feedback transformation and an output injection. Then we compare subspaces associated to a DAE in a geometric way with those associated (also in a geometric way) to a control system, namely, we compare the Wong sequences of DAEs and invariant subspaces of control systems. We prove that the KCF of linear DAEs and the MCF of control systems have a perfect correspondence and that their invariants are related. In this way, we connect the geometric analysis of linear DAEs with the classical geometric linear control theory. Finally, we propose a concept named internal equivalence for DAEs and discuss its relation with internal regularity, i.e., the existence and uniqueness of solutions.

In data-scarce scenarios, deep learning models often overfit to noise and irrelevant patterns, which limits their ability to generalize to unseen samples. To address these challenges in medical image segmentation, we introduce Diff-UMamba, a novel architecture that combines the UNet framework with the mamba mechanism to model long-range dependencies. At the heart of Diff-UMamba is a noise reduction module, which employs a signal differencing strategy to suppress noisy or irrelevant activations within the encoder. This encourages the model to filter out spurious features and enhance task-relevant representations, thereby improving its focus on clinically significant regions. As a result, the architecture achieves improved segmentation accuracy and robustness, particularly in low-data settings. Diff-UMamba is evaluated on multiple public datasets, including medical segmentation decathalon dataset (lung and pancreas) and AIIB23, demonstrating consistent performance gains of 1-3% over baseline methods in various segmentation tasks. To further assess performance under limited data conditions, additional experiments are conducted on the BraTS-21 dataset by varying the proportion of available training samples. The approach is also validated on a small internal non-small cell lung cancer dataset for the segmentation of gross tumor volume in cone beam CT, where it achieves a 4-5% improvement over baseline.

Contrastive representation learning has proven to be an effective self-supervised learning method. Most successful approaches are based on Noise Contrastive Estimation (NCE) and use different views of an instance as positives that should be contrasted with other instances, called negatives, that are considered as noise. However, several instances in a dataset are drawn from the same distribution and share underlying semantic information. A good data representation should contain relations, or semantic similarity, between the instances. Contrastive learning implicitly learns relations but considering all negatives as noise harms the quality of the learned relations. To circumvent this issue, we propose a novel formulation of contrastive learning using semantic similarity between instances called Similarity Contrastive Estimation (SCE). Our training objective is a soft contrastive learning one. Instead of hard classifying positives and negatives, we estimate from one view of a batch a continuous distribution to push or pull instances based on their semantic similarities. This target similarity distribution is sharpened to eliminate noisy relations. The model predicts for each instance, from another view, the target distribution while contrasting its positive with negatives. Experimental results show that SCE is Top-1 on the ImageNet linear evaluation protocol at 100 pretraining epochs with 72.1% accuracy and is competitive with state-of-the-art algorithms by reaching 75.4% for 200 epochs with multi-crop. We also show that SCE is able to generalize to several tasks. Source code is available here: this https URL.

The paper deals with stochastic difference-of-convex functions (DC) programs, that is, optimization problems whose the cost function is a sum of a lower semicontinuous DC function and the expectation of a stochastic DC function with respect to a probability distribution. This class of nonsmooth and nonconvex stochastic optimization problems plays a central role in many practical applications. Although there are many contributions in the context of convex and/or smooth stochastic optimization, algorithms dealing with nonconvex and nonsmooth programs remain rare. In deterministic optimization literature, the DC Algorithm (DCA) is recognized to be one of the few algorithms to solve effectively nonconvex and nonsmooth optimization problems. The main purpose of this paper is to present some new stochastic DCAs for solving stochastic DC programs. The convergence analysis of the proposed algorithms is carefully studied, and numerical experiments are conducted to justify the algorithms' behaviors.

Quantum computing leverages the principles of quantum mechanics to perform computations far beyond the capabilities of classical systems, particularly in fields such as cryptography and optimization. However, current quantum programming languages often require low-level implementation, posing significant barriers for many developers due to their steep learning curve and limited abstraction. In response, we introduce \textbf{Qutes}, a high-level quantum programming language designed to simplify quantum algorithm development while maintaining the flexibility required for advanced applications. By abstracting complex quantum operations and allowing intuitive expressions through high-level constructs, Qutes enables users to write efficient quantum programs without extensive knowledge of quantum mechanics or circuit design. Built upon Qiskit, Qutes translates its syntax directly into executable quantum code, facilitating seamless integration with quantum hardware. This paper provides an overview of the language's architecture, core functionalities, and its ability to unify classical and quantum operations within a single framework. Additionally, we demonstrate Qutes' application in key quantum algorithms, showcasing its potential to make quantum programming more accessible and practical for a wider range of developers and researchers.

Person Re-Identification (re-ID) aims at retrieving images of the same person taken by different cameras. A challenge for re-ID is the performance preservation when a model is used on data of interest (target data) which belong to a different domain from the training data domain (source data). Unsupervised Domain Adaptation (UDA) is an interesting research direction for this challenge as it avoids a costly annotation of the target data. Pseudo-labeling methods achieve the best results in UDA-based re-ID. Surprisingly, labeled source data are discarded after this initialization step. However, we believe that pseudo-labeling could further leverage the labeled source data in order to improve the post-initialization training steps. In order to improve robustness against erroneous pseudo-labels, we advocate the exploitation of both labeled source data and pseudo-labeled target data during all training iterations. To support our guideline, we introduce a framework which relies on a two-branch architecture optimizing classification and triplet loss based metric learning in source and target domains, respectively, in order to allow \emph{adaptability to the target domain} while ensuring \emph{robustness to noisy pseudo-labels}. Indeed, shared low and mid-level parameters benefit from the source classification and triplet loss signal while high-level parameters of the target branch learn domain-specific features. Our method is simple enough to be easily combined with existing pseudo-labeling UDA approaches. We show experimentally that it is efficient and improves performance when the base method has no mechanism to deal with pseudo-label noise or for hard adaptation tasks. Our approach reaches state-of-the-art performance when evaluated on commonly used datasets, Market-1501 and DukeMTMC-reID, and outperforms the state of the art when targeting the bigger and more challenging dataset MSMT.

Multimodal MR images can provide complementary information for accurate brain tumor segmentation. However, it's common to have missing imaging modalities in clinical practice. Since there exists a strong correlation between multi modalities, a novel correlation representation block is proposed to specially discover the latent multi-source correlation. Thanks to the obtained correlation representation, the segmentation becomes more robust in the case of missing modalities. The model parameter estimation module first maps the individual representation produced by each encoder to obtain independent parameters, then, under these parameters, the correlation expression module transforms all the individual representations to form a latent multi-source correlation representation. Finally, the correlation representations across modalities are fused via the attention mechanism into a shared representation to emphasize the most important features for segmentation. We evaluate our model on BraTS 2018 datasets, it outperforms the current state-of-the-art method and produces robust results when one or more modalities are missing.

Ultrafine-grained aluminum alloys offer interesting multifunctional properties with a combination of high strength, low electrical resistivity, and low density. However, due to thermally induced grain coarsening, they typically suffer from an intrinsic poor thermal stability. To overcome this drawback, an Al-2%Fe alloy has been selected because of the low solubility of Fe in Al and their highly positive enthalpy of mixing leading to the formation of stable intermetallic particles. The two-phase alloy has been processed by severe plastic deformation to achieve simultaneously submicrometer Al grains and a uniform distribution of nanoscaled intermetallic particles. The influence of the level of deformation on the microstructure has been investigated thanks to transmission electron microscopy and atom probe tomography and it is shown that for the highest strain a partial dissolution of the metastable Al6Fe particle occurred leading to the formation of a Fe super saturated solid solution. The thermal stability, and especially the precipitation of particles from the ultrafine-grained solid solution and the way they pin grain boundaries has been investigated both from static annealing and in-situ transmission electron microscopy experiments. The correlation between microstructural features and microhardness has been established to identify the various strengthening contributions. Finally, it is 2 shown that ultrafine grained high purity Al with less than 0.01 at. % Fe in solid solution could preserve a grain size only 300nm after 1h at 250$^\circ$C.

Today, deep convolutional neural networks (CNNs) have demonstrated state-of-the-art performance for medical image segmentation, on various imaging modalities and tasks. Despite early success, segmentation networks may still generate anatomically aberrant segmentations, with holes or inaccuracies near the object boundaries. To enforce anatomical plausibility, recent research studies have focused on incorporating prior knowledge such as object shape or boundary, as constraints in the loss function. Prior integrated could be low-level referring to reformulated representations extracted from the ground-truth segmentations, or high-level representing external medical information such as the organ's shape or size. Over the past few years, prior-based losses exhibited a rising interest in the research field since they allow integration of expert knowledge while still being architecture-agnostic. However, given the diversity of prior-based losses on different medical imaging challenges and tasks, it has become hard to identify what loss works best for which dataset. In this paper, we establish a benchmark of recent prior-based losses for medical image segmentation. The main objective is to provide intuition onto which losses to choose given a particular task or dataset. To this end, four low-level and high-level prior-based losses are selected. The considered losses are validated on 8 different datasets from a variety of medical image segmentation challenges including the Decathlon, the ISLES and the WMH challenge. Results show that whereas low-level prior-based losses can guarantee an increase in performance over the Dice loss baseline regardless of the dataset characteristics, high-level prior-based losses can increase anatomical plausibility as per data characteristics.

Using touch devices to navigate in virtual 3D environments such as computer assisted design (CAD) models or geographical information systems (GIS) is inherently difficult for humans, as the 3D operations have to be performed by the user on a 2D touch surface. This ill-posed problem is classically solved with a fixed and handcrafted interaction protocol, which must be learned by the user. We propose to automatically learn a new interaction protocol allowing to map a 2D user input to 3D actions in virtual environments using reinforcement learning (RL). A fundamental problem of RL methods is the vast amount of interactions often required, which are difficult to come by when humans are involved. To overcome this limitation, we make use of two collaborative agents. The first agent models the human by learning to perform the 2D finger trajectories. The second agent acts as the interaction protocol, interpreting and translating to 3D operations the 2D finger trajectories from the first agent. We restrict the learned 2D trajectories to be similar to a training set of collected human gestures by first performing state representation learning, prior to reinforcement learning. This state representation learning is addressed by projecting the gestures into a latent space learned by a variational auto encoder (VAE).

Deep Reinforcement Learning (DRL) has demonstrated impressive results in domains such as games and robotics, where task formulations are well-defined. However, few DRL benchmarks are grounded in complex, real-world environments, where safety constraints, partial observability, and the need for hand-engineered task representations pose significant challenges. To help bridge this gap, we introduce a testbed based on the pump scheduling problem in a real-world water distribution facility. The task involves controlling pumps to ensure a reliable water supply while minimizing energy consumption and respecting the constraints of the system. Our testbed includes a realistic simulator, three years of high-resolution (1-minute) operational data from human-led control, and a baseline RL task formulation. This testbed supports a wide range of research directions, including offline RL, safe exploration, inverse RL, and multi-objective optimization.

We consider the \emph{Budgeted} version of the classical \emph{Connected Dominating Set} problem (BCDS). Given a graph $G$ and a budget $k$, we seek a connected subset of at most $k$ vertices maximizing the number of dominated vertices in $G$. We improve over the previous $(1-1/e)/13$ approximation in [Khuller, Purohit, and Sarpatwar,\ \emph{SODA 2014}] by introducing a new method for performing tree decompositions in the analysis of the last part of the algorithm. This new approach provides a $(1-1/e)/12$ approximation guarantee. By generalizing the analysis of the first part of the algorithm, we are able to modify it appropriately and obtain a further improvement to $(1-e^{-7/8})/11$. On the other hand, we prove a $(1-1/e+\epsilon)$ inapproximability bound, for any $\epsilon > 0$. We also examine the \emph{edge-vertex domination} variant, where an edge dominates its endpoints and all vertices neighboring them. In \emph{Budgeted Edge-Vertex Domination} (BEVD), we are given a graph $G$, and a budget $k$, and we seek a, not necessarily connected, subset of $k$ edges such that the number of dominated vertices in $G$ is maximized. We prove there exists a $(1-1/e)$-approximation algorithm. Also, for any $\epsilon > 0$, we present a $(1-1/e+\epsilon)$-inapproximability result by a gap-preserving reduction from the \emph{maximum coverage} problem. Finally, we examine the "dual" \emph{Partial Edge-Vertex Domination} (PEVD) problem, where a graph $G$ and a quota $n'$ are given. The goal is to select a minimum-size set of edges to dominate at least $n'$ vertices in $G$. In this case, we present a $H(n')$-approximation algorithm by a reduction to the \emph{partial cover} problem.

Determinantal point processes (DPPs) have received significant attention as an elegant probabilistic model for discrete subset selection. Most prior work on DPP learning focuses on maximum likelihood estimation (MLE). While efficient and scalable, MLE approaches do not leverage any subset similarity information and may fail to recover the true generative distribution of discrete data. In this work, by deriving a differentiable relaxation of a DPP sampling algorithm, we present a novel approach for learning DPPs that minimizes the Wasserstein distance between the model and data composed of observed subsets. Through an evaluation on a real-world dataset, we show that our Wasserstein learning approach provides significantly improved predictive performance on a generative task compared to DPPs trained using MLE.

Fragility index and cooperativity length characterizing the molecular mobility in the amorphous phase are for the first time calculated in drawn polylactide (PLA). The microstructure of the samples is investigated from wide-angle X-ray scattering (WAXS) whereas the amorphous phase dynamics are revealed from broadband dielectric spectroscopy (BDS) and temperature-modulated differential scanning calorimetry (TMDSC). The drawing processes induce the decrease of both cooperativity and fragility with the orientation of the macromolecules. Post-drawing annealing reveals an unusual absence of correlation between the evolutions of cooperativity length and fragility. The cooperativity length remains the same compared to the drawn sample while a huge increase of the fragility index is recorded. By splitting the fragility index in a volume contribution and an energetic contribution, it is revealed that the amorphous phase in annealed samples exhibits a high energetic parameter, even exceeding the amorphous matrix value. It is assumed that the relaxation process is driven in such a way that the volume hindrance caused by the thermomechanical constraint is compensated by the acceleration of segmental motions linked to the increase of degrees of freedom. This result should also contribute to the understanding of the constraint slackening in the amorphous phase during annealing of drawn PLA, which causes among others the decrease of its barrier properties.

A common assumption in machine learning is that samples are independently and identically distributed (i.i.d). However, the contributions of different samples are not identical in training. Some samples are difficult to learn and some samples are noisy. The unequal contributions of samples has a considerable effect on training performances. Studies focusing on unequal sample contributions (e.g., easy, hard, noisy) in learning usually refer to these contributions as robust machine learning (RML). Weighing and regularization are two common techniques in RML. Numerous learning algorithms have been proposed but the strategies for dealing with easy/hard/noisy samples differ or even contradict with different learning algorithms. For example, some strategies take the hard samples first, whereas some strategies take easy first. Conducting a clear comparison for existing RML algorithms in dealing with different samples is difficult due to lack of a unified theoretical framework for RML. This study attempts to construct a mathematical foundation for RML based on the bias-variance trade-off theory. A series of definitions and properties are presented and proved. Several classical learning algorithms are also explained and compared. Improvements of existing methods are obtained based on the comparison. A unified method that combines two classical learning strategies is proposed.

Stochastic algorithms are well-known for their performance in the era of big data. In convex optimization, stochastic algorithms have been studied in depth and breadth. However, the current body of research on stochastic algorithms for nonsmooth, nonconvex optimization is relatively limited. In this paper, we propose new stochastic algorithms based on DC (Difference of Convex functions) programming and DCA (DC Algorithm) - the backbone of nonconvex, nonsmooth optimization. Since most real-world nonconvex programs fall into the framework of DC programming, our proposed methods can be employed in various situations, in which they confront stochastic nature and nonconvexity simultaneously. The convergence analysis of the proposed algorithms is studied intensively with the help of tools from modern convex analysis and martingale theory. Finally, we study several aspects of the proposed algorithms on an important problem in machine learning: the expected problem in Principal Component Analysis.