Search

Dynamical loss functions are derived from standard loss functions used in supervised classification tasks, but they are modified such that the contribution from each class periodically increases and decreases. These oscillations globally alter the loss landscape without affecting the global minima. In this paper, we demonstrate how to transform cross-entropy and mean squared error into dynamical loss functions. We begin by discussing the impact of increasing the size of the neural network or the learning rate on the learning process. Building on this intuition, we propose several versions of dynamical loss functions and show how they significantly improve validation accuracy for networks of varying sizes. Finally, we explore how the landscape of these dynamical loss functions evolves during training, highlighting the emergence of instabilities that may be linked to edge-of-instability minimization.

Scanning Probe Microscopy (SPM) techniques have shown great potential in fabricating nanoscale structures endowed with exotic quantum properties achieved through various manipulations of atoms and molecules. However, the selection of proper manipulation parameters requires extensive domain knowledge, which is not necessarily transferable to new systems. Therefore, efficient and autonomous SPM techniques are needed to reduce the reliance on user supervision and learn optimal strategies for new systems, in particular for the challenge of controlling chemical reactions. In this paper, we developed a software infrastructure named AutoOSS (Autonomous On-Surface Synthesis) to automate bromine removal from Zn(II)-5,15- bis(4-bromo-2,6-dimethylphenyl)porphyrin (ZnBr2Me4DPP) on Au(111), using neural network models to interpret STM outputs and deep reinforcement learning models to optimize manipulation parameters. This is further supported by Bayesian Optimization Structure Search (BOSS) and Density Functional Theory (DFT) computations to explore 3D structures and reaction mechanisms based on STM images.

Cosmic rays (CRs) have strong influences on the chemistry of dense molecular clouds (MCs). To study the detailed chemistry induced by CRs, we conducted a Yebes W band line survey towards an unshocked MC (which we named as 3C391:NML) associated with supernova remnant (SNR) 3C391. We detected emission lines of 18 molecular species in total and estimated their column densities with local thermodynamic equilibrium (LTE) and non-LTE analysis. Using the abundance ratio N(HCO+)/N(CO) and an upper limit of N(DCO+)/N(HCO+), we estimated the CR ionization rate of 3C391:NML is $\zeta\gtrsim 2.7\times 10^{-14}\rm \ s^{-1}$ with an analytic method. However, we caution on adopting this value because chemical equilibrium, which is a prerequisite of using the equations, is not necessarily reached in 3C391:NML. We observed lower N(HCO+)/N(HOC+), higher N(HCS+)/N(CS), and higher X($l$-C3H+) by an order of magnitude in 3C391:NML than the typical values in quiescent dense MCs. We found that an enhanced CR ionization rate (of order $\sim 10^{-15}$ or $\sim 10^{-14}\rm \ s^{-1}$) is needed to reproduce the observation with chemical model. This is higher than the values found in typical MCs by 2--3 orders of magnitude., Comment: 18 pages, 7 figures, accepted for publication in ApJ

Infrared dark clouds (IRDCs) are fruitful objects to study the fragmentation of interstellar filaments and initial conditions and early stages of high-mass ($M>8$ M$_{\odot}$) star formation. We used the Yebes 40 m and IRAM 30 m radio telescopes to carry out the first single-pointing spectral line observations towards the IRDC G1.75-0.08, which is a filamentary Central Molecular Zone (CMZ) cloud. Our aim is to reach an improved understanding of the gas kinematics and dynamical state of the cloud and its two clumps that we call clumps A and B. We also aim to determine the fractional abundances of the molecules detected at 3 mm towards G1.75-0.08. We detected HNCO$(J_{K_a,\,K_c}=4_{0,\,4}-3_{0,\,3})$, HCN$(J=1-0)$, and HCO$^+(J=1-0)$ towards both clumps. The N$_2$H$^+(J=1-0)$ line was detected only in clump B, while N$_2$D$^+(J=1-0)$ was not detected at all. The HCN and HNCO spectra exhibit two velocity components. The abundances of the detected species are comparable to those in other IRDCs. An upper limit to the [N$_2$D$^+$]/[N$_2$H$^+$] deuterium fraction of $<0.05$ derived towards clump B is consistent with values observed in many high-mass clumps. The line mass analysis suggests that the G1.75-0.08 filament is subcritical by a factor of $11\pm6$, and the clumps were found to be gravitationally unbound ($\alpha_{\rm vir} > 2$). Our finding that G1.75-0.08 is strongly subcritical is atypical compared to the general population of Galactic filamentary clouds. The cloud's location in the CMZ might affect the cloud kinematics similar to what has been found for the Brick IRDC, and the cloud's dynamical state might also be the result of the turbulent motions or shear and tidal forces in the CMZ. Because the target clumps are dark at 70 $\mu$m and massive (several $10^3$ M$_{\odot}$), they can be considered candidates for being high-mass starless (but not prestellar) clumps., Comment: 16 pages, 4 figures, 4 tables, accepted for publication in Galaxies, abstract abridged for arXiv

We present early results from our program of ALMA Band 6 (1.3mm) molecular line mapping of a sample of nearby, well-studied examples of high-excitation, bipolar/pinched-waist and molecule-rich planetary nebulae (Hubble 5 and NGC 2440, 2818, 2899, 6302, and 6445). We have mapped these planetary nebulae (PNe) in isotopologues of CO as well as various molecular line tracers of high-energy irradiation, such as HCN, CN, HNC, and HCO+, with the complementary goals of establishing nebular kinematics as well as the zones of UV-heated and X-ray-ionized molecular gas within each nebula. The resulting high-resolution ALMA molecular emission-line maps reveal the regions of high-excitation bipolar PNe in which molecular gas, presumably ejected during asymptotic giant branch stages of the PN progenitor stars, survives and evolves chemically. We present a summary of molecular species detected to date in the sample nebulae, and we use example results for one PN (NGC 6455) to demonstrate the power of the ALMA data in revealing the structures, kinematics, and compositions of the equatorial molecular tori that are a common feature of the sample objects., Comment: 6 pages, 4 figures, accepted for IAU Proceedings Series (IAUS 384)

Six axis force-torque sensors are commonly attached to the wrist of serial robots to measure the external forces and torques acting on the robot's end-effector. These measurements are used for load identification, contact detection, and human-robot interaction amongst other applications. Typically, the measurements obtained from the force-torque sensor are more accurate than estimates computed from joint torque readings, as the former is independent of the robot's dynamic and kinematic models. However, the force-torque sensor measurements are affected by a bias that drifts over time, caused by the compounding effects of temperature changes, mechanical stresses, and other factors. In this work, we present a pipeline that continuously estimates the bias and the drift of the bias of a force-torque sensor attached to the wrist of a robot. The first component of the pipeline is a Kalman filter that estimates the kinematic state (position, velocity, and acceleration) of the robot's joints. The second component is a kinematic model that maps the joint-space kinematics to the task-space kinematics of the force-torque sensor. Finally, the third component is a Kalman filter that estimates the bias and the drift of the bias of the force-torque sensor assuming that the inertial parameters of the gripper attached to the distal end of the force-torque sensor are known with certainty., Comment: Technical Report STARS-2024-001, University of Toronto Institute for Aerospace Studies (7 pages, 0 figure)

In this work we identify coherent electron-vibron interactions between near-resonant and non-resonant electronic levels that contribute beyond standard optomechanical models for off-resonant or resonance SERS. By developing an open-system quantum model using first molecular interaction principles, we show how the Raman interference of both resonant and non-resonant contributions can provide several orders of magnitude modifications of the SERS peaks with respect to former optomechanical models and over the fluorescence backgrounds. This cooperative optomechanical mechanism allows for generating an enhancement of nonclassical photon pair correlations between Stokes and anti-Stokes photons, which can be detected by photon-counting measurements. Our results demonstrate Raman enhancements and suppressions of coherent nature that significantly impact the standard estimations of the optomechanical contribution from SERS spectra and their quantum mechanical observable effects.

Hardware isolation primitives such as secure enclaves aim to protect sensitive programs, but remain vulnerable to transient execution attacks. Complete microarchitectural isolation is not a satisfactory defense mechanism as it leaves out public shared memory, critical for usability and application performance. Conversely, hardware-software co-designs for secure speculation can counter these attacks but are not yet practical, since they make assumptions on the speculation modes, the exposed microarchitectural state, and the software, which are all hard to support for the entire software stack. This paper advocates for processors to incorporate microarchitectural isolation primitives and mechanisms for controlled speculation, enabling different execution modes. These modes can restrict what is exposed to an attacker, effectively balancing performance and program-analysis complexity. We introduce two mechanisms to securely share memory between an enclave and an untrusted OS in an out-of-order processor. We show that our two modes are complementary, achieving speculative non-interference with a reasonable performance impact, while requiring minimal code annotation and simple program analysis doable by hand. Our prototype, Citadel, is a multicore processor running on an FPGA, booting untrusted Linux, and supporting comprehensive enclave capabilities, such as shared memory, and remote attestation. To our knowledge, Citadel is the first end-to-end enclave platform to run secure applications, such as cryptographic libraries or small private inference workloads, on a speculative out-of-order multicore processor while protecting against a significant class of side-channel attacks.

Fluid flow networks are ubiquitous and can be found in a broad range of contexts, from human-made systems such as water supply networks to living systems like animal and plant vasculature. In many cases, the elements forming these networks exhibit a highly non-linear pressure-flow relationship. Although we understand how these elements work individually, their collective behavior remains poorly understood. In this work, we combine experiments, theory, and numerical simulations to understand the main mechanisms underlying the collective behavior of soft flow networks with elements that exhibit negative differential resistance. Strikingly, our theoretical analysis and experiments reveal that a minimal network of nonlinear resistors, which we have termed a `fluidic memristor', displays history-dependent resistance. This new class of element can be understood as a collection of hysteresis loops that allows this fluidic system to store information. Our work provides insights that may inform new applications of fluid flow networks in soft materials science, biomedical settings, and soft robotics, and may also motivate new understanding of the flow networks involved in animal and plant physiology.

Planetary nebulae (PNe), the ejected envelopes of red giant stars, provide us with a history of the last, mass-losing phases of 90 percent of stars initially more massive than the Sun. Here, we analyse James Webb Space Telescope (JWST) Early Release Observation (ERO) images of the PN NGC3132. A structured, extended H2 halo surrounding an ionised central bubble is imprinted with spiral structures, likely shaped by a low-mass companion orbiting the central star at 40-60 AU. The images also reveal a mid-IR excess at the central star interpreted as a dusty disk, indicative of an interaction with another, closer companion. Including the previously known, A-type visual companion, the progenitor of the NGC3132 PN must have been at least a stellar quartet. The JWST images allow us to generate a model of the illumination, ionisation and hydrodynamics of the molecular halo, demonstrating the power of JWST to investigate complex stellar outflows. Further, new measurements of the A-type visual companion allow us to derive the value for the mass of the progenitor of a central star to date with excellent precision: 2.86+/-0.06 Mo. These results serve as path finders for future JWST observations of PNe providing unique insight into fundamental astrophysical processes including colliding winds, and binary star interactions, with implications for supernovae and gravitational wave systems., Comment: 32 pages, 5 figures for the main article. 12 pages 8 figures for the supplementary material

The young and well-studied planetary nebula NGC 7027 harbors significant molecular gas that is irradiated by luminous, point-like UV (central star) and diffuse (shocked nebular) X-ray emission. This nebula represents an excellent subject to investigate the molecular chemistry and physical conditions within photon- and X-ray-dominated regions (PDRs and XDRs). As yet, the exact formation routes of CO$^+$ and HCO$^+$ in PN environments remain uncertain. Here, we present $\sim$2$"$ resolution maps of NGC 7027 in the irradiation tracers CO$^+$ and HCO$^+$, obtained with the IRAM NOEMA interferometer, along with SMA CO and HST 2.12~$\mu$m H$_2$ data for context. The CO$^+$ map constitutes the first interferometric map of this molecular ion in any PN. Comparison of CO$^+$ and HCO$^+$ maps reveal strikingly different emission morphologies, as well as a systematic spatial displacement between the two molecules; the regions of brightest HCO$^+$, found along the central waist of the nebula, are radially offset by $\sim$1$"$ ($\sim$900 au) outside the corresponding CO$^+$ emission peaks. The CO$^+$ emission furthermore precisely traces the inner boundaries of the nebula's PDR (as delineated by near-IR H$_2$ emission), suggesting that central star UV emission drives CO$^+$ formation. The displacement of HCO$^+$ radially outward with respect to CO$^+$ is indicative that dust-penetrating soft X-rays are responsible for enhancing the HCO$^+$ abundance in the surrounding molecular envelope, forming an XDR. These interferometric CO$^+$ and HCO$^+$ observations of NGC 7027 thus clearly establish the spatial distinction between the PDR and XDR formed (respectively) by intense UV and X-ray irradiation of molecular gas., Comment: 15 pages, 7 figures, 1 table

Networks of social interactions are the substrate upon which civilizations are built. Often, we create new bonds with people that we like or feel that our relationships are damaged through the intervention of third parties. Despite their importance and the huge impact that these processes have in our lives, quantitative scientific understanding of them is still in its infancy, mainly due to the difficulty of collecting large datasets of social networks including individual attributes. In this work, we present a thorough study of real social networks of 13 schools, with more than 3,000 students and 60,000 declared positive and negative relations, including tests for personal traits of all the students. We introduce a metric -- the `triadic influence' -- that measures the influence of nearest-neighbors in the relationships of their contacts. We use neural networks to predict the relationships and to extract the probability that two students are friends or enemies depending on their personal attributes or the triadic influence. We alternatively use a high-dimensional embedding of the network structure to also predict the relationships. Remarkably, the triadic influence (a simple one-dimensional metric) achieves the highest accuracy at predicting the relationship between two students. We postulate that the probabilities extracted from the neural networks -- functions of the triadic influence and the personalities of the students -- control the evolution of real social networks, opening a new avenue for the quantitative study of these systems.

For decades, manufacturers have attempted to reduce or eliminate the optical aberrations that appear on the progressive addition lens' surfaces during manufacturing. Besides every effort made, some of these distortions are inevitable given how lenses are fabricated, where in fact, astigmatism appears on the surface and cannot be entirely removed or where non-uniform magnification becomes inherent to the power change across the lens. Some presbyopes may refer to certain discomfort when wearing these lenses for the first time, and a subset of them might never adapt. Developing, prototyping, testing and purveying those lenses into the market come at a cost, which is usually reflected in the retail price. This study aims to test the feasibility of virtual reality for testing customers' satisfaction with these lenses, even before getting them onto production. VR offers a controlled environment where different parameters affecting progressive lens comforts, such as distortions, image displacement or optical blurring, can be analysed separately. In this study, the focus was set on the distortions and image displacement, not taking blur into account. Behavioural changes (head and eye movements) were recorded using the built-in eye tracker. Participants were significantly more displeased in the presence of highly distorted lens simulations. In addition, a gradient boosting regressor was fitted to the data, so predictors of discomfort could be unveiled, and ratings could be predicted without performing additional measurements.

Active matter spans a wide range of time and length scales, from groups of cells and synthetic self-propelled particles to schools of fish, flocks of birds, or even human crowds. The theoretical framework describing these systems has shown tremendous success at finding universal phenomenology. However, further progress is often burdened by the difficulty of determining the forces that control the dynamics of the individual elements within each system. Accessing this local information is key to understanding the physics dominating the system and to create the models that can explain the observed collective phenomena. In this work, we present a machine-learning model, a graph neural network, that uses the collective movement of the system to learn the active and two-body forces controlling the individual dynamics of the particles. We verify our approach using numerical simulations of active brownian particles, considering different interaction potentials and levels of activity. Finally, we apply our model to experiments of electrophoretic Janus particles, extracting the active and two-body forces that control the dynamics of the colloids. Due to this, we can uncover the physics dominating the behavior of the system. We extract an active force that depends on the electric field and also area fraction. We also discover a dependence of the two-body interaction with the electric field that leads us to propose that the dominant force between these colloids is a screened electrostatic interaction with a constant length scale. We expect that this methodology can open a new avenue for the study and modeling of experimental systems of active particles.

Observations of molecular clouds, prestellar cores, and protoplanetary disks have established that the HNC/HCN ratio may be a potent diagnostic of molecular gas physical conditions. The processes that govern the relative abundances of these molecules nevertheless remain poorly understood. We seek to exploit the wide range of UV irradiation strengths within the 1 pc diameter Helix planetary nebula to explore the potential role of UV radiation in driving HNC/HCN. We performed IRAM 30 m and APEX 12 m radio line observations across six positions within the Helix Nebula, making use of radiative transfer and photodissociation modeling codes to interpret the results for line intensities and line ratios in terms of the molecular gas properties. We have obtained the first detections of the plasma-embedded Helix molecular knots (globules) in HCN, HNC, HCO+, and other trace molecules. Analysis of the HNC/HCN integrated line intensity ratio reveals an increase with radial distance from the Helix central star. In the context of molecular line ratios of other planetary nebulae from the literature, the HNC/HCN ratio appears to be anticorrelated with UV emission over four orders of magnitude in incident flux. Models of the photodissociation regions within the Helix using the RADEX and Meudon codes reveal strong constraints on column density of the molecular gas, as well as pressure and temperature. Analysis of the molecular ion HCO+ across the Helix indicates that X-ray irradiation is likely driving HCO+ production in the outer regions of planetary nebulae, where photodissociation is limited, yet cold gas and ionized molecules are abundant. Although the observational results clearly indicate that UV irradiation is important in determining the HNC/HCN ratio, our PDR modeling indicates that the UV flux gradient alone cannot reproduce the observed variation of HNC/HCN across the Helix., Comment: 19 pages, 13 figures, 6 tables

Flow networks efficiently transport nutrients and other solutes in many physical systems, such as plant and animal vasculature. In the case of the animal circulatory system, an adequate oxygen and nutrient supply is not guaranteed everywhere: as nutrients travel through the microcirculation and get absorbed, they become less available at the venous side of the vascular network. Ensuring that the nutrient distribution is homogeneous provides a fitness advantage, as all tissue gets enough supply to survive while waste is minimized. How do animals build such a uniform perfusing flow system? We propose a local adaptation rule for the vessel radii that is able to equalize perfusion, while minimizing energy dissipation to circulate the flow and the material cost. The rule is a combination of different objective cost functions that compete to produce complex network morphologies ranging from hierarchical architectures to uniform mesh grids, depending on how each cost is weighted. We find that our local adaptation rules are consistent with experimental data of the rat mesentery vasculature., Comment: 9 pages, 5 figures