Search

The management of natural environments, whether for conservation or production, requires a deep understanding of wildlife. The number, location, and behavior of wild animals are among the main subjects of study in ecology and wildlife research. The use of camera traps offers the opportunity to quickly collect large quantities of photographs that capture wildlife in its natural habitat, avoiding factors that could alter their behavior. In Tierra del Fuego, Argentina, research is being conducted on forest use by different herbivores (guanacos, cows, sheep) to optimize management and protect these natural ecosystems. Although camera traps allow for the collection of millions of images, interpreting such photographs presents a scalability challenge for manual processing. As a result, much of the valuable knowledge stored in these vast data repositories remains untapped. Neural Networks and Deep Learning are areas of study within Artificial Intelligence. Over the past decade, these two disciplines have made significant contributions to image recognition on a global scale. Ecological and wildlife conservation studies can be combined with these new technologies to extract important information from the photographs obtained by camera traps, contributing to the understanding of various natural processes and improving the management of the involved wild areas. Our project aims to develop neural network models to classify animal species in photographs taken with camera traps, addressing large-scale challenges in scientific research., Comment: in Spanish language, XXIV Workshop de Investigadores en Ciencias de la Computaci\'on (WICC 2022, Mendoza)

This paper presents a novel deep learning-based approach named RealDiffFusionNet incorporating Neural Controlled Differential Equations (Neural CDE) - time series models that are robust in handling irregularly sampled data - and multi-head attention to align relevant multimodal context (image data, time invariant data, etc.) at each time point. Long short-term memory (LSTM) models were also used as a baseline. Two different datasets were used: a data from the Open-Source Imaging Consortium (OSIC) containing structured time series data of demographics and lung function with a baseline CT scan of the lungs and the second from the Alzheimer's Disease Neuroimaging Initiative (ADNI) containing a series of MRI scans along with demographics, physical examinations, and cognitive assessment data. An ablation study was performed to understand the role of CDEs, multimodal data, attention fusion, and interpolation strategies on model performance. When the baseline models were evaluated, the use of multimodal data resulted in an improvement in Neural CDE performance, with a lower test RMSE. Additionally, the performance of multimodal Neural CDE was also superior to multimodal LSTM. In the attention-based architectures, fusion through concatenation and rectilinear interpolation were found to improve model performance. The performance of the proposed RealDiffFusionNet was found to be superior (0.2570) to all models. For the ADNI dataset, between the Neural-CDE and LSTM models trained only on the structured data, the test RMSE were comparable (0.471 for LSTM vs. 0.4581 Neural-CDE). Furthermore, the addition of image features from patients' MRI series resulted in an improvement in performance, with a lower test RMSE (0.4372 with multimodal vs 0.4581 with structured data). RealDiffFusionNet has shown promise in utilizing CDEs and multimodal data to accurately predict disease progression.

We discuss $(K,N)$-convexity and gradient flows for $(K,N)$-convex functionals on metric spaces, in the case of real $K$ and negative $N$. In this generality, it is necessary to consider functionals unbounded from below and/or above, possibly attaining as values both the positive and the negative infinity. We prove several properties of gradient flows of $(K,N)$-convex functionals characterized by Evolution Variational Inequalities, including contractivity, regularity, and uniqueness., Comment: 23 pages

Measuring fluctuations in matter's low energy excitations is the key to unveil the nature of the nonequilibrium response of materials. A promising outlook in this respect is offered by spectroscopic methods that address matter fluctuations by exploiting the statistical nature of light-matter interactions with weak few-photon probes. Here we report the first implementation of ultrafast phase randomized tomography, combining pump-probe experiments with quantum optical state tomography, to measure the ultrafast non-equilibrium dynamics in complex materials. Our approach utilizes a time-resolved multimode heterodyne detection scheme with phase-randomized coherent ultrashort laser pulses, overcoming the limitations of phase-stable configurations and enabling a robust reconstruction of the statistical distribution of phase-averaged optical observables. This methodology is validated by measuring the coherent phonon response in $\alpha$-quartz. By tracking the dynamics of the shot-noise limited photon number distribution of few-photon probes with ultrafast resolution, our results set an upper limit to the non-classical features of phononic state in $\alpha$-quartz and provide a pathway to access nonequilibrium quantum fluctuations in more complex quantum materials., Comment: 21 pages, 10 figures (including Supplementary Information)

The use of optical cavities on resonance with material excitations allows controlling light-matter interaction in both the regimes of weak and strong coupling. We study here the multimode vibrational coupling of low energy phonons in the charge-density-wave material 1T-TaS$_{2}$ across its insulator-to-metal phase transition. For this purpose, we embed 1T-TaS$_{2}$ into THz Fabry-P\'erot cryogenic cavities tunable in frequency within the spectral range of the vibrational modes of the insulating phase and track the linear response of the coupled phonons across the insulator-to-metal transition. In the low temperature dielectric state, we reveal the signatures of a multimode vibrational strong collective coupling. The observed polariton modes inherit character from all the vibrational resonances as a consequence of the cavity-mediated hybridization. We reveal that the vibrational strong collective coupling is suppressed across the insulator-to-metal transition as a consequence of the phonon-screening induced by the free charges. Our findings emphasize how the response of cavity-coupled vibrations can be modified by the presence of free charges, uncovering a new direction toward the tuning of coherent light-matter interaction in cavity-confined correlated materials., Comment: 24 pages, 9 figures (including Supplementary Material)

There has been substantial growth in the UAV market along with an expansion in their applications. However, the successful execution of a UAV mission is very often dependent on the use of a GNSS. Unfortunately, the vulnerability of GNSS signals, due to their lack of encryption and authentication, poses a significant cybersecurity issue. This vulnerability makes various attacks, particularly the "GNSS spoofing attack," and "GNSS jamming attack" easily executable. Generally speaking, during this attack, the drone is manipulated into altering its path, usually resulting in an immediate forced landing or crash. As far as we know, we are the first to propose a lightweight-solution that enable a drone to autonomously rescue itself, assuming it is under GNSS attack and the GNSS is no longer available, and return safely to its initial takeoff position, thereby preventing any potential crashes. During the flight, wind plays a critical role as it can instantaneously alter the drone's position. To solve this problem, we have devised a highly effective 2-phases solution: (i) Forward Phase, for monitoring and recording the forward journey, and (ii) Backward Phase, that generates a backward route, based on the Forward Phase and wind presence. The final solution ensures strong performance in consistently returning the drone to the original position, even in wind situations, while maintaining a very fast computation time., Comment: This article has been accepted for publication in the proceedings of the 2024 IEEE International Conference on Wireless and Mobile Computing, Networking And Communications (WiMob)

We study the convergence of an $N$-particle Markovian controlled system to the solution of a family of stochastic McKean-Vlasov control problems, either with a finite horizon or Schr\"odinger type cost functional. Specifically, under suitable assumptions, we prove the convergence of the value functions, the fixed-time probability distributions, and the relative entropy of their path-space probability laws. These proofs are based on a Benamou-Brenier type reformulation of the problem and a superposition principle, both of which are tools from the theory of optimal transport., Comment: 54 pages. Comments are welcome!

The analysis of complex high-dimensional data is a common task in many domains, resulting in bespoke visual exploration tools. Expectations and practices of domain experts as users do not always align with visualization theory. In this paper, we report on a design study in the medical domain where we developed two high-fidelity prototypes encoding EEG-derived brain network data with different types of visualizations. We evaluate these prototypes regarding effectiveness, efficiency, and preference with two groups: participants with domain knowledge (domain experts in medical research) and those without domain knowledge, both groups having little or no visualization experience. A requirement analysis and study of low-fidelity prototypes revealed a strong preference for a novel and aesthetically pleasing visualization design, as opposed to a design that is considered more optimal based on visualization theory. Our study highlights the pros and cons of both approaches, discussing trade-offs between task-specific measurements and subjective preference. While the aesthetically pleasing and novel low-fidelity prototype was favored, the results of our evaluation show that, in most cases, this was not reflected in participants' performance or subjective preference for the high-fidelity prototypes., Comment: 13 pages, 4 figures, 2 tables

Inspired by emergent behaviors of living matter, there is increasing interest in developing approaches to create dynamic patterns and structures in synthetic materials with controllable complexity to enable functionalities that are not possible in thermodynamic equilibrium. Here we show that electrophoretically driven and maintained non-equilibrium gradients of magnetic nanoparticles in non-polar solvent can undergo electrokinetic instabilities (EKI), leading to various electrically controllable spatiotemporally patterned states. These electrokinetic instabilities and patterns can be tuned with a magnetic field via magnetostatic energy reduction mechanism to both increase and decrease the pattern complexity. We reflect the experimental observations on the theoretical electrokinetic and magnetostatic arguments. We further show that small amounts of polar water in the otherwise non-polar system are critical enablers for the electrophoretic mobility of the nanoparticles. Since functionalities of magnetic nanoparticles are widely tunable, we foresee that the combination of dissipative electrokinetic driving and magnetic energy reduction can lead to novel functional dissipative materials.

Multi-tasking machine learning (ML) models exhibit prediction abilities in domains with little to no training data available (few-shot and zero-shot learning). Over-parameterized ML models are further capable of zero-loss training and near-optimal generalization performance. An open research question is, how these novel paradigms contribute to solving tasks related to enhancing the renewable energy transition and mitigating climate change. A collection of unified ML tasks and datasets from this domain can largely facilitate the development and empirical testing of such models, but is currently missing. Here, we introduce the ETT-17 (Energy Transition Tasks-17), a collection of 17 datasets from six different application domains related to enhancing renewable energy, including out-of-distribution validation and testing data. We unify all tasks and datasets, such that they can be solved using a single multi-tasking ML model. We further analyse the dimensions of each dataset; investigate what they require for designing over-parameterized models; introduce a set of dataset scores that describe important properties of each task and dataset; and provide performance benchmarks.

We establish a nonlinear analogue of a splitting map into a Euclidean space, as a harmonic map into a flat torus. We prove that the existence of such a map implies Gromov-Hausdorff closeness to a flat torus in any dimension. Furthermore, Gromov-Hausdorff closeness to a flat torus and an integral bound {on $r_M(x)$, the smallest eigenvalue of the Ricci tensor $\text{ric}_x$ in $x$}, imply the existence of a harmonic splitting map. Combining these results with Stern's inequality, we provide a new Gromov-Hausdorff stability theorem for flat $3$-tori. The main tools we employ include the harmonic map heat flow, Ricci flow, and both Ricci limits and RCD theories., Comment: 44 pages. We corrected a mistake in Theorem 5.3, present in the previous version, and added Theorem 1.8

In the cuprates, the opening of a d-wave superconducting (SC) gap is accompanied by a redistribution of spectral weight at energies two orders of magnitude larger than this gap. This indicates the importance to the pairing mechanism of on-site electronic excitations, such as orbital transitions or charge transfer excitations. Here, we resort to a three-pulse pump-probe scheme to study the broadband non-equilibrium dielectric function in optimally-doped $Bi_2Sr_2Ca_{0.92}Y_{0.08}Cu_2O_{8+{\delta}}$ and we identify two distinct dynamical responses: i) a blueshift of the central energy of an interband excitation peaked at 2 eV and ii) a change in spectral weight in the same energy range. Photoexcitation with near-IR and mid-IR pulses, with photon energies respectively above and below the SC gap, reveals that the transient changes in the central energy are not modified by the onset of superconductivity and do not depend on the pump photon energy. Conversely, the spectral weight dynamics strongly depends on the pump photon energy and has a discontinuity at the critical temperature. The picture that emerges is that, while high-energy pulses excite quasiparticles in both nodal and thermally inaccessible antinodal states, photoexcitation by low-energy pulses mostly accelerates the condensate and creates excitations predominantly at the nodes of the SC gap. These results, rationalized by kinetic equations for d-wave superconducting gaps, indicate that dynamical control of the momentum-dependent distribution of non-thermal quasiparticles may be achieved by the selective tuning of the photoexcitation.

For negative ion beam sources there are several methods of measuring the accelerated beam current, most commonly electrical measurements at the power supply and calorimetric measurements. On SPIDER, the ITER Heating Neutral Beam full-scale beam source prototype, electrical measurements at the acceleration grid power supply (AGPS) are complemented by polarizing the diagnostic calorimeter STRIKE to provide an additional electrical measurement of the accelerated current. This is in addition to the calorimetric measurements provided by STRIKE. These diagnostics give differing measurements of the beam current. Exploiting the reduced number of open apertures on SPIDER a new beam diagnostic has been installed to measure the individual beamlet currents directly. The so called Beamlet Current Monitor (BCM) has been used to measure the current of five beamlets during the most recent SPIDER campaign. This work compares the BCM current to the electrical measurements at the AGPS and STRIKE. The average BCM current agrees well with the STRIKE electrical measurements, indicating that the AGPS overestimates the beam current. The individual beamlets are compared to the STRIKE calorimetric measurements, showing similar current trends with the source parameters.

Identifying molecules that exhibit some pre-specified properties is a difficult problem to solve. In the last few years, deep generative models have been used for molecule generation. Deep Graph Variational Autoencoders are among the most powerful machine learning tools with which it is possible to address this problem. However, existing methods struggle in capturing the true data distribution and tend to be computationally expensive. In this work, we propose RGCVAE, an efficient and effective Graph Variational Autoencoder based on: (i) an encoding network exploiting a new powerful Relational Graph Isomorphism Network; (ii) a novel probabilistic decoding component. Compared to several state-of-the-art VAE methods on two widely adopted datasets, RGCVAE shows state-of-the-art molecule generation performance while being significantly faster to train.

Using only image-sentence pairs, weakly-supervised visual-textual grounding aims to learn region-phrase correspondences of the respective entity mentions. Compared to the supervised approach, learning is more difficult since bounding boxes and textual phrases correspondences are unavailable. In light of this, we propose the Semantic Prior Refinement Model (SPRM), whose predictions are obtained by combining the output of two main modules. The first untrained module aims to return a rough alignment between textual phrases and bounding boxes. The second trained module is composed of two sub-components that refine the rough alignment to improve the accuracy of the final phrase-bounding box alignments. The model is trained to maximize the multimodal similarity between an image and a sentence, while minimizing the multimodal similarity of the same sentence and a new unrelated image, carefully selected to help the most during training. Our approach shows state-of-the-art results on two popular datasets, Flickr30k Entities and ReferIt, shining especially on ReferIt with a 9.6% absolute improvement. Moreover, thanks to the untrained component, it reaches competitive performances just using a small fraction of training examples.

SPIDER is the full-scale prototype of the ion source of the ITER Heating Neutral Beam Injector, where negative ions of Hydrogen or Deuterium are produced by a RF generated plasma and accelerated with a set of grids up to ~100 keV. The Power Supply System is composed of high voltage dc power supplies capable of handling frequent grid breakdowns, high current dc generators for the magnetic filter field and RF generators for the plasma generation. During the first 3 years of SPIDER operation different electrical issues were discovered, understood and addressed thanks to deep analyses of the experimental results supported by modelling activities. The paper gives an overview on the observed phenomena and relevant analyses to understand them, on the effectiveness of the short-term modifications provided to SPIDER to face the encountered issues and on the design principle of long-term solutions to be introduced during the currently ongoing long shutdown., Comment: 8 pages, 12 figures. Presented at SOFT 2022

ITER envisages the use of two heating neutral beam injectors plus an optional one as part of the auxiliary heating and current drive system. The 16.5 MW expected neutral beam power per injector is several notches higher than worldwide existing facilities. A Neutral Beam Test Facility (NBTF) was established at Consorzio RFX, exploiting the synergy of two test beds, SPIDER and MITICA. SPIDER is dedicated to developing and characterizing large efficient negative ion sources at relevant parameters in ITER-like conditions: source and accelerator located in the same vacuum where the beam propagates, immunity to electromagnetic interferences of multiple radio-frequency (RF) antennas, avoidance of RF-induced discharges on the outside of the source. Three years of experiments on SPIDER have addressed to the necessary design modifications to enable full performances. The source is presently under a long shut-down phase to incorporate learnings from the experimental campaign. Parallelly, developments on MITICA, the full-scale prototype of the ITER NBI featuring a 1 MV accelerator and ion neutralization, are underway including manufacturing of in-vessel components, while power supplies and auxiliary plants are already under final testing and commissioning. Integration, commissioning and tests of the 1MV power supplies are essential for this first-of-kind system, unparalleled both in research and industry field. The integrated test to confirm 1MV output by combining invertor systems, DC generators and transmission lines extracted errors/accidents in some components. To realize a concrete system for ITER, solutions for the repair and the improvement of the system were developed. Hence, NBTF is emerging as a necessary facility, due to the large gap with existing injectors, effectively dedicated to identify issues and find solutions to enable successful ITER NBI operations in a time bound fashion.

Placing quantum materials into optical cavities provides a unique platform for controlling quantum cooperative properties of matter, via both weak and strong light-matter coupling. Here we report the experimental evidence of reversible cavity control of a metal-to-insulator phase transition in a correlated solid-state material. We embed the charge density wave material 1T-TaS$_{2}$ into cryogenic tunable terahertz cavities and show that a switch between conductive and insulating behaviors, associated with a large change in the sample temperature, is obtained by mechanically tuning the distance between the cavity mirrors and their alignment. The large thermal modification observed is indicative of a Purcell-like scenario in which the spectral profile of the cavity modifies the energy exchange between the material and the external electromagnetic field. Our findings provide opportunities for controlling the thermodynamics and macroscopic transport properties of quantum materials by engineering their electromagnetic environment., Comment: 52 pages, 36 figures (including Supplementary Information)

The aim of this paper is to show the existence of a canonical distance $\mathsf d'$ defined on a locally Minkowski metric measure space $(\mathsf X,\mathsf d,\mathfrak m)$ such that: i) $\mathsf d'$ is equivalent to $\mathsf d$, ii) $(\mathsf X, \mathsf d', \mathfrak m)$ is infinitesimally Hilbertian. This new regularity assumption on $(\mathsf X, \mathsf d,\mathfrak m)$ essentially forces the structure to be locally similar to a Minkowski space and defines a class of metric measure structures which includes all the Finsler manifolds, and it is actually strictly larger. The required distance $\mathsf d'$ will be the intrinsic distance $\mathsf d_\mathsf{KS}$ associated to the so-called Korevaar-Schoen energy, which is proven to be a quadratic form. In particular, we show that the Cheeger energy associated to the metric measure space $(\mathsf X, \mathsf d_\mathsf{KS}, \mathfrak m)$ is in fact the Korevaar-Schoen energy.