Search

Recent advancements in Text-to-Speech (TTS) technology have led to natural-sounding speech for English, primarily due to the availability of large-scale, high-quality web data. However, many other languages lack access to such resources, relying instead on limited studio-quality data. This scarcity results in synthesized speech that often suffers from intelligibility issues, particularly with low-frequency character bigrams. In this paper, we propose three solutions to address this challenge. First, we leverage high-quality data from linguistically or geographically related languages to improve TTS for the target language. Second, we utilize low-quality Automatic Speech Recognition (ASR) data recorded in non-studio environments, which is refined using denoising and speech enhancement models. Third, we apply knowledge distillation from large-scale models using synthetic data to generate more robust outputs. Our experiments with Hindi demonstrate significant reductions in intelligibility issues, as validated by human evaluators. We propose this methodology as a viable alternative for languages with limited access to high-quality data, enabling them to collectively benefit from shared resources., Comment: 11 pages, 1 figure, 3 tables

Recent advancements in text-to-speech (TTS) synthesis show that large-scale models trained with extensive web data produce highly natural-sounding output. However, such data is scarce for Indian languages due to the lack of high-quality, manually subtitled data on platforms like LibriVox or YouTube. To address this gap, we enhance existing large-scale ASR datasets containing natural conversations collected in low-quality environments to generate high-quality TTS training data. Our pipeline leverages the cross-lingual generalization of denoising and speech enhancement models trained on English and applied to Indian languages. This results in IndicVoices-R (IV-R), the largest multilingual Indian TTS dataset derived from an ASR dataset, with 1,704 hours of high-quality speech from 10,496 speakers across 22 Indian languages. IV-R matches the quality of gold-standard TTS datasets like LJSpeech, LibriTTS, and IndicTTS. We also introduce the IV-R Benchmark, the first to assess zero-shot, few-shot, and many-shot speaker generalization capabilities of TTS models on Indian voices, ensuring diversity in age, gender, and style. We demonstrate that fine-tuning an English pre-trained model on a combined dataset of high-quality IndicTTS and our IV-R dataset results in better zero-shot speaker generalization compared to fine-tuning on the IndicTTS dataset alone. Further, our evaluation reveals limited zero-shot generalization for Indian voices in TTS models trained on prior datasets, which we improve by fine-tuning the model on our data containing diverse set of speakers across language families. We open-source all data and code, releasing the first TTS model for all 22 official Indian languages., Comment: Accepted to NeurIPS 2024 Datasets and Benchmarks track

3D generative models have shown significant promise in structure-based drug design (SBDD), particularly in discovering ligands tailored to specific target binding sites. Existing algorithms often focus primarily on ligand-target binding, characterized by binding affinity. Moreover, models trained solely on target-ligand distribution may fall short in addressing the broader objectives of drug discovery, such as the development of novel ligands with desired properties like drug-likeness, and synthesizability, underscoring the multifaceted nature of the drug design process. To overcome these challenges, we decouple the problem into molecular generation and property prediction. The latter synergistically guides the diffusion sampling process, facilitating guided diffusion and resulting in the creation of meaningful molecules with the desired properties. We call this guided molecular generation process as TAGMol. Through experiments on benchmark datasets, TAGMol demonstrates superior performance compared to state-of-the-art baselines, achieving a 22% improvement in average Vina Score and yielding favorable outcomes in essential auxiliary properties. This establishes TAGMol as a comprehensive framework for drug generation.

We describe the symmetry group of the stabilizer polytope for any number $n$ of systems and any prime local dimension $d$. In the qubit case, the symmetry group coincides with the linear and anti-linear Clifford operations. In the case of qudits, the structure is somewhat richer: for $n=1$, it is a wreath product of permutations of bases and permutations of the elements within each basis. For $n>1$, the symmetries are given by affine symplectic similitudes. These are the affine maps that preserve the symplectic form of the underlying discrete phase space up to a non-zero multiplier. We phrase these results with respect to a number of a priori different notions of "symmetry'', including Kadison symmetries (bijections that are compatible with convex combinations), Wigner symmetries (bijections that preserve inner products), and symmetries realized by an action on Hilbert space. Going beyond stabilizer states, we extend an observation of Heinrich and Gross (Ref. [25]) and show that the symmetries of fairly general sets of Hermitian operators are constrained by certain moments. In particular: the symmetries of a set that behaves like a 3-design preserve Jordan products and are therefore realized by conjugation with unitaries or anti-unitaries. (The structure constants of the Jordan algebra are encoded in an order-three tensor, which we connect to the third moments of a design). This generalizes Kadison's formulation of the classic Wigner Theorem on quantum mechanical symmetries., Comment: 21 pages, v2: minor notation changes and references added

The dishonest casino is a well-known hidden Markov model (HMM) used in educational settings to introduce HMMs and graphical models. Here, a sequence of die rolls is observed, with the casino switching between a fair and a loaded die. Typically, the goal is to use the observed rolls to infer the pattern of fair and loaded dice, leading to filtering, smoothing, and Viterbi algorithms. This paper, however, explores how much of the winnings is attributable to the casino's cheating, a counterfactual question beyond the scope of HMM primitives. To address this, we introduce a structural causal model (SCM) consistent with the HMM and show that the expected winnings attributable to cheating (EWAC) can be bounded using linear programs (LPs). Through numerical experiments, we compute these bounds and develop intuition using benchmark SCMs based on independence, comonotonic, and counter-monotonic copulas. We show that tighter bounds are obtained with a time-homogeneity condition on the SCM, while looser bounds allow for an almost explicit LP solution. Domain-specific knowledge like pathwise monotonicity or counterfactual stability can be incorporated via linear constraints. Our work contributes to bounding counterfactuals in causal inference and is the first to develop LP bounds in a dynamic HMM setting, benefiting educational contexts where counterfactual inference is taught., Comment: arXiv admin note: text overlap with arXiv:2205.13832

In the e-commerce domain, the accurate extraction of attribute-value pairs from product listings (e.g., Brand: Apple) is crucial for enhancing search and recommendation systems. The automation of this extraction process is challenging due to the vast diversity of product categories and their respective attributes, compounded by the lack of extensive, accurately annotated training datasets and the demand for low latency to meet the real-time needs of e-commerce platforms. To address these challenges, we introduce GenToC, a novel two-stage model for extracting attribute-value pairs from product titles. GenToC is designed to train with partially-labeled data, leveraging incomplete attribute-value pairs and obviating the need for a fully annotated dataset. Moreover, we introduce a bootstrapping method that enables GenToC to progressively refine and expand its training dataset. This enhancement substantially improves the quality of data available for training other neural network models that are typically faster but are inherently less capable than GenToC in terms of their capacity to handle partially-labeled data. By supplying an enriched dataset for training, GenToC significantly advances the performance of these alternative models, making them more suitable for real-time deployment. Our results highlight the unique capability of GenToC to learn from a limited set of labeled data and to contribute to the training of more efficient models, marking a significant leap forward in the automated extraction of attribute-value pairs from product titles. GenToC has been successfully integrated into India's largest B2B e-commerce platform, IndiaMART.com, achieving a significant increase of 21.1% in recall over the existing deployed system while maintaining a high precision of 89.5% in this challenging task.

Despite the rapid advancement in the field of image recognition, the processing of high-resolution imagery remains a computational challenge. However, this processing is pivotal for extracting detailed object insights in areas ranging from autonomous vehicle navigation to medical imaging analyses. Our study introduces a framework aimed at mitigating these challenges by leveraging memory efficient patch based processing for high resolution images. It incorporates a global context representation alongside local patch information, enabling a comprehensive understanding of the image content. In contrast to traditional training methods which are limited by memory constraints, our method enables training of ultra high resolution images. We demonstrate the effectiveness of our method through superior performance on 7 different benchmarks across classification, object detection, and segmentation. Notably, the proposed method achieves strong performance even on resource-constrained devices like Jetson Nano. Our code is available at https://github.com/Visual-Conception-Group/Localized-Perception-Constrained-Vision-Systems.

Sky distributions of large samples of distant active galactic nuclei (AGNs) have shown dipoles significantly larger than the cosmic microwave background (CMB) dipole. However, a recent Bayesian analysis of the QUAIA sample, comprising 1.3 million quasars, has yielded a dipole that seems to be in tandem with the CMB dipole, in contravention of most previous studies of AGN dipoles. Since the question has large cosmological implications, we investigate the QUAIA quasar sample afresh, by directly computing the dipole from asymmetries observed in the source number counts. We instead find a dipole 3-4 times as large as the CMB dipole though in the same direction. Further, it has been claimed elsewhere that the difference between the CMB dipole and the radio dipole estimated from the NRAO VLA Sky Survey (NVSS), the first large catalogue that showed an AGN dipole about four times larger than the CMB dipole, can be fully accounted for by incorporating the shot-noise and clustering contributions to the total NVSS dipole. A careful reinvestigation of the NVSS dipole, however, shows that the random phenomena like shot noise or clustering cannot account for the actually observed NVSS asymmetries, which show a systematic dipole pattern over the sky., Comment: 6 pages, 6 figures, 2 tables, accepted in MNRAS

Chess graphs encode the moves that a particular chess piece can make on an $m\times n$ chessboard. We study through these graphs through the lens of chip-firing games and graph gonality. We provide upper and lower bounds for the gonality of king's, bishop's, and knight's graphs, as well as for the toroidal versions of these graphs. We also prove that among all chess graphs, there exists an upper bound on gonality solely in terms of $\min\{m,n\}$, except for queen's, toroidal queen's, rook's, and toroidal bishop's graphs., Comment: 24 pages, 25 figures

Solid-state detectors with a low energy threshold have several applications, including in direct-detection searches of non-relativistic halo dark-matter particles with sub-GeV masses. Moreover, when searching for relativistic or quasi-relativistic beyond-the-Standard-Model particles (i.e., $v/c\gtrsim 0.01$) that have an enhanced cross section for small energy transfers, a comparatively small detector with a low energy threshold may have better sensitivity than a larger detector with a higher energy threshold. In this paper, we provide accurate calculations of the low-energy ionization spectrum from high-velocity particles scattering in a dielectric material. We focus on silicon, although our results can be easily applied to other materials. We consider the full material response, in particular also the excitation of bulk plasmons. We generalize the energy-loss function to relativistic kinematics, and benchmark existing tools used for halo dark-matter scattering against publicly available electron energy-loss spectroscopy data. Compared to calculations of energy loss that are commonly used in the literature, such as the Photo-Absorption-Ionization model or the free-electron model, the inclusion of collective effects shifts the recoil ionization spectrum towards higher energies, typically peaking around 4--6 electron-hole pairs. We apply our results to the three benchmark examples: millicharged particles produced in a beam, neutrinos with a magnetic dipole moment produced in a reactor, and dark-matter particles that are upscattered by cosmic rays or in the Sun. Our results show that the proper inclusion of collective effects typically enhances a detector's sensitivity to these particles, since detector backgrounds, such as dark counts, peak at lower energies., Comment: 26 pages, 9 figures including appendices and references

The electromagnetic energy-momentum of a moving charged spherical capacitor may be calculated by a 4-vector Lorentz transformation from the energy in the rest frame. However, energy-momentum of the moving system computed directly from electromagnetic fields yields extra terms; in particular a factor of 4/3 in momentum appears, similar to that encountered in the classical electron model, where this enigmatic factor has been a source of scourge for more than a century. There have been many attempts to eliminate this `unwanted' factor, noteworthy among them is a modification in electromagnetic field energy-momentum definition that has entered even standard textbooks. Here it is shown that in a moving charged spherical capacitor, some additional contributions to the electromagnetic energy-momentum arise naturally from electromagnetic forces in system or equivalently from terms in the Maxwell stress tensor; contributions that do not otherwise show up in 4-vector transformations. Furthermore, a similar factor of 4/3 in the momentum of a perfect fluid comprising a randomly moving ultra-relativistic gas molecules or an isotropic photon gas, filling an {\em uncharged} spherical capacitor in motion, appears owing to the contribution of pressure. Thus, genesis of the "enigmatic" factor of 4/3 can be traced to pressure or stress whose presence in the system may be of non-electromagnetic origin and where the proposed modifications in energy-momentum definition do not even come into picture. No modifications in the definitions of energy-momentum of moving fluids have ever been required; physics should be the same in electromagnetic case as well, implying there is nothing amiss in the standard definition of electromagnetic energy-momentum., Comment: 10 pages, 3 figures

The flexibility of choosing the ad action as a function of the consumer state is critical for modern-day marketing campaigns. We study the problem of identifying the optimal sequential personalized interventions that maximize the adoption probability for a new product. We model consumer behavior by a conversion funnel that captures the state of each consumer (e.g., interaction history with the firm) and allows the consumer behavior to vary as a function of both her state and firm's sequential interventions. We show our model captures consumer behavior with very high accuracy (out-of-sample AUC of over 0.95) in a real-world email marketing dataset. However, it results in a very large-scale learning problem, where the firm must learn the state-specific effects of various interventions from consumer interactions. We propose a novel attribution-based decision-making algorithm for this problem that we call model-free approximate Bayesian learning. Our algorithm inherits the interpretability and scalability of Thompson sampling for bandits and maintains an approximate belief over the value of each state-specific intervention. The belief is updated as the algorithm interacts with the consumers. Despite being an approximation to the Bayes update, we prove the asymptotic optimality of our algorithm and analyze its convergence rate. We show that our algorithm significantly outperforms traditional approaches on extensive simulations calibrated to a real-world email marketing dataset.

We present the first results from a dark matter search using six Skipper-CCDs in the SENSEI detector operating at SNOLAB. With an exposure of 534.9 gram-days from well-performing sensors, we select events containing 2 to 10 electron-hole pairs. After aggressively masking images to remove backgrounds, we observe 55 two-electron events, 4 three-electron events, and no events containing 4 to 10 electrons. The two-electron events are consistent with pileup from one-electron events. Among the 4 three-electron events, 2 appear in pixels that are likely impacted by detector defects, although not strongly enough to trigger our "hot-pixel" mask. We use these data to set world-leading constraints on sub-GeV dark matter interacting with electrons and nuclei., Comment: 5 pages, 2 figures, + Supplemental Materials (5 pages, 5 figures) + References

We investigate dipole asymmetries in four large radio surveys, each spanning more than 80\% of the sky. Two of them, the Very Large Array Sky Survey (VLASS) and the Rapid ASKAP Continuum Survey (RACS), have recently yielded dipoles that appear incongruent with each other as well as seem inconsistent with previous radio survey dipoles and the Cosmic Microwave Background (CMB) dipole. Because these radio surveys have large overlaps in sky coverage, comprising hence large majority of common radio sources, one would not expect significant differences between their radio dipoles, irrespective of their underlying source of origin. We examine here in detail these radio dipoles, to ascertain the source of incongruency amongst them. We find the VLASS and RACS data to be containing some declination-dependent systematics, seemingly in the vicinity of the declination limit of each survey. We show that the effects of such systematics can be mitigated by restricting the declination limits of the respective survey during the dipole determination. A weighted mean of the sky coordinates of thus derived dipoles from the four radio surveys lies within $1.2\sigma$ of the CMB dipole direction. However, the amplitude appears significantly larger, $3.7\pm 0.6$ times or more than the CMB dipole. This puts in doubt not only the conventional wisdom that the genesis of all these dipoles, including that of the CMB dipole, is due to the Solar peculiar motion, it also raises uncomfortable questions about the Cosmological Principle (CP), the basis of the standard $\Lambda$CDM cosmological model., Comment: 15 pages, 8 figures, 7 tables, accepted in MNRAS

The origin of the radio synchrotron background (RSB) is currently unknown. Its understanding might have profound implications in fundamental physics or might reveal a new class of radio emitters. In this work, we consider the scenario in which the RSB is due to extragalactic radio sources and measure the angular cross-correlation of LOFAR images of the diffuse radio sky with matter tracers at different redshifts, provided by galaxy catalogs and CMB lensing. We compare these measured cross-correlations to those expected for models of RSB sources. We find that low-redshift populations of discrete sources are excluded by the data, while higher redshift explanations are compatible with available observations. We also conclude that at least 20\% of the RSB surface brightness level must originate from populations tracing the large-scale distribution of matter in the universe, indicating that at least this fraction of the RSB is of extragalactic origin. Future measurements of the correlation between the RSB and tracers of high-redshift sources will be crucial to constraining the source population of the RSB., Comment: 12 pages, 8 figures. Version accepted for publication in MNRAS

Low-energy Compton scattering is an important background for sub-GeV dark matter direct-detection and other experiments. Current Compton scattering calculations typically rely on assumptions that are not valid in the low-energy region of interest, beneath ~ 50 eV. Here we relate the low-energy Compton scattering differential cross section to the dielectric response of the material. Our new approach can be used for a wide range of materials and includes all-electron, band structure, and collective effects, which can be particularly relevant at low energies. We demonstrate the strength of our approach in several solid-state systems, in particular Si, Ge, GaAs, and SiC, which are relevant for current and proposed experiments searching for dark matter, neutrinos, and millicharged particles., Comment: 5 figures, 9 pages + references

We present Project Florida, a system architecture and software development kit (SDK) enabling deployment of large-scale Federated Learning (FL) solutions across a heterogeneous device ecosystem. Federated learning is an approach to machine learning based on a strong data sovereignty principle, i.e., that privacy and security of data is best enabled by storing it at its origin, whether on end-user devices or in segregated cloud storage silos. Federated learning enables model training across devices and silos while the training data remains within its security boundary, by distributing a model snapshot to a client running inside the boundary, running client code to update the model, and then aggregating updated snapshots across many clients in a central orchestrator. Deploying a FL solution requires implementation of complex privacy and security mechanisms as well as scalable orchestration infrastructure. Scale and performance is a paramount concern, as the model training process benefits from full participation of many client devices, which may have a wide variety of performance characteristics. Project Florida aims to simplify the task of deploying cross-device FL solutions by providing cloud-hosted infrastructure and accompanying task management interfaces, as well as a multi-platform SDK supporting most major programming languages including C++, Java, and Python, enabling FL training across a wide range of operating system (OS) and hardware specifications. The architecture decouples service management from the FL workflow, enabling a cloud service provider to deliver FL-as-a-service (FLaaS) to ML engineers and application developers. We present an overview of Florida, including a description of the architecture, sample code, and illustrative experiments demonstrating system capabilities.

We calculate target-material responses for dark matter--electron scattering at the \textit{ab-initio} all-electron level using atom-centered gaussian basis sets. The all-electron effects enhance the material response at high momentum transfers from dark matter to electrons, $q\gtrsim \mathcal{O}\left({10\ \alpha m_e}\right)$, compared to calculations using conventional plane wave methods, including those used in QEDark; this enhances the expected event rates at energy transfers $E \gtrsim 10$~eV, especially when scattering through heavy mediators. We carefully test a range of systematic uncertainties in the theory calculation, including those arising from the choice of basis set, exchange-correlation functional, number of unit cells in the Bloch sum, $\mathbf{k}$-mesh, and neglect of scatters with very high momentum transfers. We provide state-of-the-art crystal form factors, focusing on silicon and germanium. Our code and results are made publicly available as a new tool, called Quantum Chemistry Dark (``QCDark'')., Comment: 15 pages, plus references and appendices, 12 figures

We present results exploring the role that probabilistic deep learning models can play in cosmology from large-scale astronomical surveys through photometric redshift (photo-z) estimation. Photo-z uncertainty estimates are critical for the science goals of upcoming large-scale surveys such as LSST, however common machine learning methods typically provide only point estimates and lack uncertainties on predictions. We turn to Bayesian neural networks (BNNs) as a promising way to provide accurate predictions of redshift values with uncertainty estimates. We have compiled a galaxy data set from the Hyper Suprime-Cam Survey with grizy photometry, which is designed to be a smaller scale version of large surveys like LSST. We use this data set to investigate the performance of a neural network (NN) and a probabilistic BNN for photo-z estimation and evaluate their performance with respect to LSST photo-z science requirements. We also examine the utility of photo-z uncertainties as a means to reduce catastrophic outlier estimates. The BNN outputs the estimate in the form of a Gaussian probability distribution. We use the mean and standard deviation as the redshift estimate and uncertainty. We find that the BNN can produce accurate uncertainties. Using a coverage test, we find excellent agreement with expectation -- 67.2$\%$ of galaxies between $0 < 2.5$ have 1-$\sigma$ uncertainties that cover the spectroscopic value. We also include a comparison to alternative machine learning models using the same data. We find the BNN meets two out of three of the LSST photo-z science requirements in the range $0 < z < 2.5$., Comment: 17 pages, 23 figures, 4 tables