Search

Humans excel at building generalizations of new concepts from just one single example. Contrary to this, current computer vision models typically require large amount of training samples to achieve a comparable accuracy. In this work we present a Bayesian model of perception that learns using only minimal data, a prototypical probabilistic program of an object. Specifically, we propose a generative inverse graphics model of primitive shapes, to infer posterior distributions over physically consistent parameters from one or several images. We show how this representation can be used for downstream tasks such as few-shot classification and pose estimation. Our model outperforms existing few-shot neural-only classification algorithms and demonstrates generalization across varying lighting conditions, backgrounds, and out-of-distribution shapes. By design, our model is uncertainty-aware and uses our new differentiable renderer for optimizing global scene parameters through gradient descent, sampling posterior distributions over object parameters with Markov Chain Monte Carlo (MCMC), and using a neural based likelihood function.

Automatic Speech Recognition (ASR) or Speech-to-text (STT) has greatly evolved in the last few years. Traditional architectures based on pipelines have been replaced by joint end-to-end (E2E) architectures that simplify and streamline the model training process. In addition, new AI training methods, such as weak-supervised learning have reduced the need for high-quality audio datasets for model training. However, despite all these advancements, little to no research has been done on real-time transcription. In real-time scenarios, the audio is not pre-recorded, and the input audio must be fragmented to be processed by the ASR systems. To achieve real-time requirements, these fragments must be as short as possible to reduce latency. However, audio cannot be split at any point as dividing an utterance into two separate fragments will generate an incorrect transcription. Also, shorter fragments provide less context for the ASR model. For this reason, it is necessary to design and test different splitting algorithms to optimize the quality and delay of the resulting transcription. In this paper, three audio splitting algorithms are evaluated with different ASR models to determine their impact on both the quality of the transcription and the end-to-end delay. The algorithms are fragmentation at fixed intervals, voice activity detection (VAD), and fragmentation with feedback. The results are compared to the performance of the same model, without audio fragmentation, to determine the effects of this division. The results show that VAD fragmentation provides the best quality with the highest delay, whereas fragmentation at fixed intervals provides the lowest quality and the lowest delay. The newly proposed feedback algorithm exchanges a 2-4% increase in WER for a reduction of 1.5-2s delay, respectively, to the VAD splitting., Comment: 15 pages, 4 figures

The 16 Myr-old A0V star HD 156623 in the Scorpius--Centaurus association hosts a high-fractional-luminosity debris disk, recently resolved in scattered light for the first time by the Gemini Planet Imager (GPI) in polarized intensity. We present new analysis of the GPI H-band polarimetric detection of the HD 156623 debris disk, with particular interest in its unique morphology. This debris disk lacks a visible inner clearing, unlike the majority of low-inclination disks in the GPI sample and in Sco-Cen, and it is known to contain CO gas, positioning it as a candidate ``hybrid'' or ``shielded'' disk. We use radiative transfer models to constrain the geometric parameters of the disk based on scattered light data and thermal models to constrain the unresolved inner radius based on the system's spectral energy distribution (SED). We also compute a measurement of the polarized scattering phase function, adding to the existing sample of empirical phase function measurements. We find that HD 156623's debris disk inner radius is constrained to less than 26.6 AU from scattered light imagery and less than 13.4 AU from SED modeling at a 99.7% confidence interval, and suggest that gas drag may play a role in retaining sub-blowout size dust grains so close to the star., Comment: 18 pages, 10 figures. Accepted for publication in AJ

There is growing recognition among financial institutions, financial regulators and policy makers of the importance of addressing nature-related risks and opportunities. Evaluating and assessing nature-related risks for financial institutions is challenging due to the large volume of heterogeneous data available on nature and the complexity of investment value chains and the various components' relationship to nature. The dual problem of scaling data analytics and analysing complex systems can be addressed using Artificial Intelligence (AI). We address issues such as plugging existing data gaps with discovered data, data estimation under uncertainty, time series analysis and (near) real-time updates. This report presents potential AI solutions for models of two distinct use cases, the Brazil Beef Supply Use Case and the Water Utility Use Case. Our two use cases cover a broad perspective within sustainable finance. The Brazilian cattle farming use case is an example of greening finance - integrating nature-related considerations into mainstream financial decision-making to transition investments away from sectors with poor historical track records and unsustainable operations. The deployment of nature-based solutions in the UK water utility use case is an example of financing green - driving investment to nature-positive outcomes. The two use cases also cover different sectors, geographies, financial assets and AI modelling techniques, providing an overview on how AI could be applied to different challenges relating to nature's integration into finance. This report is primarily aimed at financial institutions but is also of interest to ESG data providers, TNFD, systems modellers, and, of course, AI practitioners., Comment: 67 pages, 10 figures, UKRI (NERC) Integrated Finance and Biodiversity for a Nature Positive Future Programme

In this paper, we explore the challenges of ensuring security and privacy for users from diverse demographic backgrounds. We propose a threat modeling approach to identify potential risks and countermeasures for product inclusion in security and privacy. We discuss various factors that can affect a user's ability to achieve a high level of security and privacy, including low-income demographics, poor connectivity, shared device usage, ML fairness, etc. We present results from a global security and privacy user experience survey and discuss the implications for product developers. Our work highlights the need for a more inclusive approach to security and privacy and provides a framework for researchers and practitioners to consider when designing products and services for a diverse range of users.

The surface tension of monolayers with mixtures of anionic and nonionic surfactant at the liquid/vapour interface is studied. Previous works have observed that calculations of the surface tension of simple fluids show artificial oscillations for small interfacial areas, indicating that the surface tension data fluctuate due to the finite size effects and periodic boundary conditions. In the case of simulations of monolayers composed of surfactant mixtures, the surface tension not only oscillates for small areas but can also give non-physical data, such as negative values. Analysis of the monolayers with different surfactant mixtures, ionic (DTAB, CTAB, SDS) and nonionic (SB3-12), was done for density profiles, parameters of order and pair correlation functions for small and large box areas and all of them present similar behaviour. The fluctuations and the non-physical values of the surface tension are corrected when boxes with large interfacial areas are considered. The results indicate that in order to obtain reliable values of the surface tension, in computer simulations, it is important to choose not only the correct force field but also the appropriate size of the simulation box., Comment: 14 pages, 9 figures, 6 tables

In many developing countries, access to electricity remains a significant challenge. Electrification planners in these countries often have to make important decisions on the mode of electrification and the planning of electrical networks for those without access, while under resource constraints. An integrated approach to electrification planning in which traditional grid electrification is complemented off-the-grid technologies such as off-grid microgrids and stand-alone systems can enable the economic provision of electricity access in these regions. This integrated planning approach can be facilitated by determining the least-cost mode of electrification - i.e by electric grid extension or off-grid systems - for non-electrified consumers in a region under analysis, while considering technical, economic and environmental constraints. Computational clustering methods the identification of consumer clusters (either as clusters of off-grid microgrids, stand-alone systems or grid-extension projects) can be undertaken using computational clustering methods. This paper presents a novel computational approach to achieve this purpose. This methodology involves exploiting the grid network that connects all consumers, by greedily partitioning the network to identify clusters of consumers to be electrified by grid-extension and off-grid microgrid systems. Using test cases and sensitivity analyses, we implement and benchmark this top-down approach with those obtained from a bottom-up clustering methodology used by the Reference Electrification Model, a model obtainable in literature. Results presented show that the alternative top-down methodology proposed can compare favorably, in terms of global electrification costs, with a bottom-up approach to rural electrification planning.

In this work, we obtain closed expressions for the transfer matrix and the transmittance of electromagnetic waves propagating in finite 1D anisotropic periodic stratified media with an arbitrary number of cells. By invoking the Cayley-Hamilton theorem on the transfer matrix for the electromagnetic field in a periodic stratified media formed by N cells, we obtain a fourth-degree recursive relation for the matrix coefficients that defines the so-called Tetranacci Polynomials. In the symmetric case, corresponding to a unit-cell transfer matrix with a characteristic polynomial where the coefficients of the linear and cubic terms are equal, closed expressions for the solutions to the recursive relation, known as symmetric Tetranacci Polynomials, have recently been derived, allowing us to write the transfer matrix and transmittance in a closed form. We show as sufficient conditions that the $4\times4$ differential propagation matrix of each layer in the binary unit cell, $\Delta$, a) has eigenvalues of the form $\pm p_1$, $\pm p_2$, with $p_1\ne p_2$, and b) its off-diagonal $2\times2$ block matrices possess the same symmetric structure in both layers. Otherwise, the recursive relations are still solvable for any $4\times4$-matrix and provide an algorithm to compute the N-th power of the transfer matrix without carrying out explicitly the matrix multiplication of N matrices. We obtain analytical expressions for the dispersion relation and transmittance, in closed form, for two finite periodic systems: the first one consists of two birefringent uniaxial media with their optical axis perpendicular to the z-axis, and the second consists of two isotropic media subject to an external magnetic field oriented along the z-axis and exhibiting the Faraday effect. Our formalism applies also to lossy media, magnetic anisotropy or optical activity., Comment: 36 pages, 6 figures, submitted to Journal of Physics A: Mathematical and Theoretical

The interaction of surfaces in relative motion in wet environments is dominated by lubrication forces, which play a pivotal role in the dynamics of microscopic systems. Here, we develop motile vesicles that exploit lubrication forces to roll on substrates. The activity of the vesicle comes from the confined rotational flow generated by a driven rotating particle encapsulated within the vesicle by droplet-microfluidics. Lubrication forces driving vesicle rolling are controlled by membrane mechanics and its tribological properties. This provides the design principles for motile vesicles that exploit frictional forces to efficiently navigate through complex environments., Comment: 6 pages, 5 figures

The Gemini Planet Imager (GPI) has excelled in imaging debris disks in the near-infrared. The GPI Exoplanet Survey (GPIES) imaged twenty-four debris disks in polarized $H$-band light, while other programs observed half of these disks in polarized $J$- and/or $K1$-bands. Using these data, we present a uniform analysis of the morphology of each disk to find asymmetries suggestive of perturbations, particularly those due to planet-disk interactions. The multi-wavelength surface brightness, the disk color and geometry permit identification of any asymmetries such as warps or disk offsets from the central star. We find that nineteen of the disks in this sample exhibit asymmetries in surface brightness, disk color, disk geometry, or a combination of the three, suggesting that for this sample, perturbations, as seen in scattered light, are common. The relationship between these perturbations and potential planets in the system are discussed. We also explore correlations among stellar temperatures, ages, disk properties, and observed perturbations. We find significant trends between the vertical aspect ratio and the stellar temperature, disk radial extent, and the dust grain size distribution power-law, $q$. We also confirm a trend between the disk color and stellar effective temperature, where the disk becomes increasingly red/neutral with increasing temperature. Such results have important implications on the evolution of debris disk systems around stars of various spectral types., Comment: 46 pages, 20 figures, 6 tables, accepted for publication in ApJ

Inspired by the structure of top-down derived models endowed with modular flavor symmetries, we investigate the yet phenomenologically unexplored binary dihedral group 2D_3. After building the vector-valued modular forms in the representations of 2D_3 with small modular weights, we systematically classify all (Dirac and Majorana) mass textures of fermions with fractional modular weights and all possible 2+1-family structures. This allows us to explore the parameter space of fermion models based on 2D_3, aiming at a description of both quarks and leptons with a minimal number of parameters and best compatibility with observed data. We consider the separate possibilities of neutrino masses generated by either a type-I seesaw mechanism or the Weinberg operator. We identify a model that, besides fitting all known flavor observables, delivers predictions for six not-yet measured parameters and favors normal-ordered neutrino masses generated by the Weinberg operator. It would be interesting to figure out whether it is possible to embed our model within a top-down scheme, such as T2/Z4 heterotic orbifold compactifications., Comment: 33 pages + references, 9 tables. v2: new discussion on a solution to the mu problem and references added

Photovoltaics (PVs) are a critical technology for curbing growing levels of anthropogenic greenhouse gas emissions, and meeting increases in future demand for low-carbon electricity. In order to fulfil ambitions for net-zero carbon dioxide equivalent (CO2eq) emissions worldwide, the global cumulative capacity of solar PVs must increase by an order of magnitude from 0.9 TWp in 2021 to 8.5 TWp by 2050 according to the International Renewable Energy Agency, which is considered to be a highly conservative estimate. In 2020, the Henry Royce Institute brought together the UK PV community to discuss the critical technological and infrastructure challenges that need to be overcome to address the vast challenges in accelerating PV deployment. Herein, we examine the key developments in the global community, especially the progress made in the field since this earlier roadmap, bringing together experts primarily from the UK across the breadth of the photovoltaics community. The focus is both on the challenges in improving the efficiency, stability and levelized cost of electricity of current technologies for utility-scale PVs, as well as the fundamental questions in novel technologies that can have a significant impact on emerging markets, such as indoor PVs, space PVs, and agrivoltaics. We discuss challenges in advanced metrology and computational tools, as well as the growing synergies between PVs and solar fuels, and offer a perspective on the environmental sustainability of the PV industry. Through this roadmap, we emphasize promising pathways forward in both the short- and long-term, and for communities working on technologies across a range of maturity levels to learn from each other., Comment: 160 pages, 21 figures

The presence of a population of a large number ($\sim$400) of almost coeval (100--300 Myr) super star clusters (SSCs) in the disk of M82 offers an opportunity to construct the Cluster Initial Mass Function (CIMF) from the observed present-day Cluster Mass Function (CMF). We carry out the dynamical and photometric evolution of the CMF assuming the clusters move in circular orbits under the gravitational potential of the host galaxy using the semi-analytical simulation code EMACSS. We explore power-law and log-normal functions for the CIMFs, and populate the clusters in the disk assuming uniform, power-law, and exponential radial distribution functions. We find that the observed CMF is best produced by a CIMF that is power-law in form with an index of 1.8, for a power-law radial distribution function. More importantly, we establish that the observed turn-over in the present-day CMF is the result of observational incompleteness rather than due to dynamically induced effects, or an intrinsically log-normal CIMF, as was proposed for the fossil starburst region B of this galaxy. Our simulations naturally reproduce the mass-radius relation observed for a sub-sample of M82 SSCs., Comment: 17 pages, 11 figures, accepted to be published on MNRAS

Saliency methods are frequently used to explain Deep Neural Network-based models. Adebayo et al.'s work on evaluating saliency methods for classification models illustrate certain explanation methods fail the model and data randomization tests. However, on extending the tests for various state of the art object detectors we illustrate that the ability to explain a model is more dependent on the model itself than the explanation method. We perform sanity checks for object detection and define new qualitative criteria to evaluate the saliency explanations, both for object classification and bounding box decisions, using Guided Backpropagation, Integrated Gradients, and their Smoothgrad versions, together with Faster R-CNN, SSD, and EfficientDet-D0, trained on COCO. In addition, the sensitivity of the explanation method to model parameters and data labels varies class-wise motivating to perform the sanity checks for each class. We find that EfficientDet-D0 is the most interpretable method independent of the saliency method, which passes the sanity checks with little problems., Comment: 18 pages, 10 figures, 1st World Conference on eXplainable Artificial Intelligence camera ready

The applicability of the three steps systematic parametrization procedure (3SSPP) to develop a force field for primary amines was evaluated in the present work. Previous simulations of primary amines show that current force fields (FF) can underestimate some experimental values under room conditions. Therefore, we propose a new set of parameters, for an united atom (UA) model, that can be used for short and long amines which predict correctly thermodynamic and dynamical properties. Following the 3SSPP methodology, the partial charges are chosen to match the experimental dielectric constant whereas the Lennard-Jones (LJ) parameters, $\epsilon$ and $\sigma$, are fitted to reproduce the surface tension at the vapor-liquid interface and the liquid density, respectively. Simulations were initially conducted for the propylamine molecule by introducing three different types of carbon atoms, C$_\alpha$ and C$_\beta$, with electric charges, and C$_n$, without charge. Then, modifying the charges of the carbons and using the transferable LJ parameters, the new set of constants for long amines were found. The results show good agreement for the experimental dielectric constant and mass density with a percentage error less than 1%, whereas for the surface tension the error is up to 4%. For the short amines, methylamine and ethylamine, the new charges were obtained from a fitting function calculated from the long amines results. For these molecules, the values of the dielectric constant and the surface tension present errors of the order of 10% with the experimental data. Miscibility of the amines was also tested with the new parameters and the results show reasonable agreement with experiments., Comment: 13 pages, 9 figures, 4 tables

We report progress on calculations of the heavy-light baryons $\Sigma_c$ and $\Lambda_c$ and their excitations with $J^P = 1/2^+$ using functional methods. We employ a covariant quark-diquark approach, where the interaction amounts to a quark exchange between quarks and diquarks and the ingredients are determined from the quark level. A partial-wave analysis reveals the presence of orbital angular momentum components in terms of p waves, which are non-relativistically suppressed., Comment: 9 pages, 5 figures, Proceedings of conference Baryons 2022 -- International Conference on the Structure of Baryons

State-of-the-art object detectors are treated as black boxes due to their highly non-linear internal computations. Even with unprecedented advancements in detector performance, the inability to explain how their outputs are generated limits their use in safety-critical applications. Previous work fails to produce explanations for both bounding box and classification decisions, and generally make individual explanations for various detectors. In this paper, we propose an open-source Detector Explanation Toolkit (DExT) which implements the proposed approach to generate a holistic explanation for all detector decisions using certain gradient-based explanation methods. We suggests various multi-object visualization methods to merge the explanations of multiple objects detected in an image as well as the corresponding detections in a single image. The quantitative evaluation show that the Single Shot MultiBox Detector (SSD) is more faithfully explained compared to other detectors regardless of the explanation methods. Both quantitative and human-centric evaluations identify that SmoothGrad with Guided Backpropagation (GBP) provides more trustworthy explanations among selected methods across all detectors. We expect that DExT will motivate practitioners to evaluate object detectors from the interpretability perspective by explaining both bounding box and classification decisions., Comment: 24 pages, with appendix. 1st World Conference on eXplainable Artificial Intelligence camera ready