Search

In this paper, we study the determination of Hamiltonian from a given equations of motion. It can be cast into a problem of matrix factorization after reinterpretation of the system as first order evolutionary equations in the phase space coordinates. We state the criterion on the evolution matrix for a Hamiltonian to exist. In addition, the proof is constructive and an explicit Hamiltonian with accompanied symplectic structure can be obtained. As an application, we will study a class of PT-symmetric systems for illustration.

Automatic Speech Assessment (ASA) has seen notable advancements with the utilization of self-supervised features (SSL) in recent research. However, a key challenge in ASA lies in the imbalanced distribution of data, particularly evident in English test datasets. To address this challenge, we approach ASA as an ordinal classification task, introducing Weighted Vectors Ranking Similarity (W-RankSim) as a novel regularization technique. W-RankSim encourages closer proximity of weighted vectors in the output layer for similar classes, implying that feature vectors with similar labels would be gradually nudged closer to each other as they converge towards corresponding weighted vectors. Extensive experimental evaluations confirm the effectiveness of our approach in improving ordinal classification performance for ASA. Furthermore, we propose a hybrid model that combines SSL and handcrafted features, showcasing how the inclusion of handcrafted features enhances performance in an ASA system., Comment: Accepted to Interspeech 2024

Self-play via online learning is one of the premier ways to solve large-scale two-player zero-sum games, both in theory and practice. Particularly popular algorithms include optimistic multiplicative weights update (OMWU) and optimistic gradient-descent-ascent (OGDA). While both algorithms enjoy $O(1/T)$ ergodic convergence to Nash equilibrium in two-player zero-sum games, OMWU offers several advantages including logarithmic dependence on the size of the payoff matrix and $\widetilde{O}(1/T)$ convergence to coarse correlated equilibria even in general-sum games. However, in terms of last-iterate convergence in two-player zero-sum games, an increasingly popular topic in this area, OGDA guarantees that the duality gap shrinks at a rate of $O(1/\sqrt{T})$, while the best existing last-iterate convergence for OMWU depends on some game-dependent constant that could be arbitrarily large. This begs the question: is this potentially slow last-iterate convergence an inherent disadvantage of OMWU, or is the current analysis too loose? Somewhat surprisingly, we show that the former is true. More generally, we prove that a broad class of algorithms that do not forget the past quickly all suffer the same issue: for any arbitrarily small $\delta>0$, there exists a $2\times 2$ matrix game such that the algorithm admits a constant duality gap even after $1/\delta$ rounds. This class of algorithms includes OMWU and other standard optimistic follow-the-regularized-leader algorithms., Comment: 27 pages, 4 figures

Speech segmentation is an essential part of speech translation (ST) systems in real-world scenarios. Since most ST models are designed to process speech segments, long-form audio must be partitioned into shorter segments before translation. Recently, data-driven approaches for the speech segmentation task have been developed. Although the approaches improve overall translation quality, a performance gap exists due to a mismatch between the models and ST systems. In addition, the prior works require large self-supervised speech models, which consume significant computational resources. In this work, we propose a segmentation model that achieves better speech translation quality with a small model size. We propose an ASR-with-punctuation task as an effective pre-training strategy for the segmentation model. We also show that proper integration of the speech segmentation model into the underlying ST system is critical to improve overall translation quality at inference time., Comment: Accepted to Interspeech 2024

We introduce a sequence of orthogonal polynomials whose associated moments are the Rayleigh-type sums, involving the zeros of the Bessel derivative $J_\nu'$ of order $\nu$. We also discuss the fundamental properties of those polynomials such as recurrence, orthogonality, etc. Consequently, we obtain a formula for the Hankel determinant, elements of which are chosen as the aforementioned Rayleigh-type sums. As an application, we complete the Hurwitz-type theorem for $J_\nu'$, which deals with the number of complex zeros of $J_\nu'$ depending on the range of $\nu$., Comment: 33 pages, 4 figures

The Spoken Language Understanding Evaluation (SLUE) suite of benchmark tasks was recently introduced to address the need for open resources and benchmarking of complex spoken language understanding (SLU) tasks, including both classification and sequence generation tasks, on natural speech. The benchmark has demonstrated preliminary success in using pre-trained speech foundation models (SFM) for these SLU tasks. However, the community still lacks a fine-grained understanding of the comparative utility of different SFMs. Inspired by this, we ask: which SFMs offer the most benefits for these complex SLU tasks, and what is the most effective approach for incorporating these SFMs? To answer this, we perform an extensive evaluation of multiple supervised and self-supervised SFMs using several evaluation protocols: (i) frozen SFMs with a lightweight prediction head, (ii) frozen SFMs with a complex prediction head, and (iii) fine-tuned SFMs with a lightweight prediction head. Although the supervised SFMs are pre-trained on much more speech recognition data (with labels), they do not always outperform self-supervised SFMs; the latter tend to perform at least as well as, and sometimes better than, supervised SFMs, especially on the sequence generation tasks in SLUE. While there is no universally optimal way of incorporating SFMs, the complex prediction head gives the best performance for most tasks, although it increases the inference time. We also introduce an open-source toolkit and performance leaderboard, SLUE-PERB, for these tasks and modeling strategies., Comment: Accepted at ACL Findings 2024

We present the Black Hole Explorer (BHEX), a mission that will produce the sharpest images in the history of astronomy by extending submillimeter Very-Long-Baseline Interferometry (VLBI) to space. BHEX will discover and measure the bright and narrow "photon ring" that is predicted to exist in images of black holes, produced from light that has orbited the black hole before escaping. This discovery will expose universal features of a black hole's spacetime that are distinct from the complex astrophysics of the emitting plasma, allowing the first direct measurements of a supermassive black hole's spin. In addition to studying the properties of the nearby supermassive black holes M87* and Sgr A*, BHEX will measure the properties of dozens of additional supermassive black holes, providing crucial insights into the processes that drive their creation and growth. BHEX will also connect these supermassive black holes to their relativistic jets, elucidating the power source for the brightest and most efficient engines in the universe. BHEX will address fundamental open questions in the physics and astrophysics of black holes that cannot be answered without submillimeter space VLBI. The mission is enabled by recent technological breakthroughs, including the development of ultra-high-speed downlink using laser communications, and it leverages billions of dollars of existing ground infrastructure. We present the motivation for BHEX, its science goals and associated requirements, and the pathway to launch within the next decade., Comment: Proceedings for SPIE Astronomical Telescopes and Instrumentation

AMoRE-II aims to search for neutrinoless double beta decay with an array of 423 Li$_2$$^{100}$MoO$_4$ crystals operating in the cryogenic system as the main phase of the Advanced Molybdenum-based Rare process Experiment (AMoRE). AMoRE has been planned to operate in three phases: AMoRE-pilot, AMoRE-I, and AMoRE-II. AMoRE-II is currently being installed at the Yemi Underground Laboratory, located approximately 1000 meters deep in Jeongseon, Korea. The goal of AMoRE-II is to reach up to $T^{0\nu\beta\beta}_{1/2}$ $\sim$ 6 $\times$ 10$^{26}$ years, corresponding to an effective Majorana mass of 15 - 29 meV, covering all the inverted mass hierarchy regions. To achieve this, the background level of the experimental configurations and possible background sources of gamma and beta events should be well understood. We have intensively performed Monte Carlo simulations using the GEANT4 toolkit in all the experimental configurations with potential sources. We report the estimated background level that meets the 10$^{-4}$counts/(keV$\cdot$kg$\cdot$yr) requirement for AMoRE-II in the region of interest (ROI) and show the projected half-life sensitivity based on the simulation study.

Maximum Manifold Capacity Representations (MMCR) is a recent multi-view self-supervised learning (MVSSL) method that matches or surpasses other leading MVSSL methods. MMCR is intriguing because it does not fit neatly into any of the commonplace MVSSL lineages, instead originating from a statistical mechanical perspective on the linear separability of data manifolds. In this paper, we seek to improve our understanding and our utilization of MMCR. To better understand MMCR, we leverage tools from high dimensional probability to demonstrate that MMCR incentivizes alignment and uniformity of learned embeddings. We then leverage tools from information theory to show that such embeddings maximize a well-known lower bound on mutual information between views, thereby connecting the geometric perspective of MMCR to the information-theoretic perspective commonly discussed in MVSSL. To better utilize MMCR, we mathematically predict and experimentally confirm non-monotonic changes in the pretraining loss akin to double descent but with respect to atypical hyperparameters. We also discover compute scaling laws that enable predicting the pretraining loss as a function of gradients steps, batch size, embedding dimension and number of views. We then show that MMCR, originally applied to image data, is performant on multimodal image-text data. By more deeply understanding the theoretical and empirical behavior of MMCR, our work reveals insights on improving MVSSL methods.

Previous foundation models for retinal images were pre-trained with limited disease categories and knowledge base. Here we introduce RetiZero, a vision-language foundation model that leverages knowledge from over 400 fundus diseases. To RetiZero's pre-training, we compiled 341,896 fundus images paired with text descriptions, sourced from public datasets, ophthalmic literature, and online resources, encompassing a diverse range of diseases across multiple ethnicities and countries. RetiZero exhibits superior performance in several downstream tasks, including zero-shot disease recognition, image-to-image retrieval, and internal- and cross-domain disease identification. In zero-shot scenarios, RetiZero achieves Top5 accuracy scores of 0.8430 for 15 fundus diseases and 0.7561 for 52 fundus diseases. For image retrieval, it achieves Top5 scores of 0.9500 and 0.8860 for the same disease sets, respectively. Clinical evaluations show that RetiZero's Top3 zero-shot performance surpasses the average of 19 ophthalmologists from Singapore, China and the United States. Furthermore, RetiZero significantly enhances clinicians' accuracy in diagnosing fundus disease. These findings underscore the value of integrating the RetiZero foundation model into clinical settings, where a variety of fundus diseases are encountered.

This work proposes an efficient method to enhance the quality of corrupted speech signals by leveraging both acoustic and visual cues. While existing diffusion-based approaches have demonstrated remarkable quality, their applicability is limited by slow inference speeds and computational complexity. To address this issue, we present FlowAVSE which enhances the inference speed and reduces the number of learnable parameters without degrading the output quality. In particular, we employ a conditional flow matching algorithm that enables the generation of high-quality speech in a single sampling step. Moreover, we increase efficiency by optimizing the underlying U-net architecture of diffusion-based systems. Our experiments demonstrate that FlowAVSE achieves 22 times faster inference speed and reduces the model size by half while maintaining the output quality. The demo page is available at: https://cyongong.github.io/FlowAVSE.github.io/, Comment: INTERSPEECH 2024

ARM Memory Tagging Extension (MTE) is a new hardware feature introduced in ARMv8.5-A architecture, aiming to detect memory corruption vulnerabilities. The low overhead of MTE makes it an attractive solution to mitigate memory corruption attacks in modern software systems and is considered the most promising path forward for improving C/C++ software security. This paper explores the potential security risks posed by speculative execution attacks against MTE. Specifically, this paper identifies new TikTag gadgets capable of leaking the MTE tags from arbitrary memory addresses through speculative execution. With TikTag gadgets, attackers can bypass the probabilistic defense of MTE, increasing the attack success rate by close to 100%. We demonstrate that TikTag gadgets can be used to bypass MTE-based mitigations in real-world systems, Google Chrome and the Linux kernel. Experimental results show that TikTag gadgets can successfully leak an MTE tag with a success rate higher than 95% in less than 4 seconds. We further propose new defense mechanisms to mitigate the security risks posed by TikTag gadgets.

The Choi-state is an indispensable tool in the study and analysis of quantum channels. Considering a channel in terms of its associated Choi-state can greatly simplify problems. It also offers an alternative approach to the characterisation of a channel, with properties of the Choi-state providing novel insight into a channel's behaviour. The rank of a Choi-state, termed the Choi-rank, has proven to be an important characterising property, and here, its significance is further elucidated through an operational interpretation. The Choi-rank is shown to provide a universal bound on how successfully two agents, Alice and Bob, can perform an entanglement-assisted exclusion task. The task can be considered an extension of super-dense coding, where Bob can only output information about Alice's encoded bit-string with certainty. Conclusive state exclusion, in place of state discrimination, is therefore considered at the culmination of the super-dense coding protocol. In order to prove this result, a necessary condition for conclusive k-state exclusion of a set of states is presented in order to achieve this result, and the notions of weak and strong exclusion are introduced., Comment: 6 + 3 pages, 2 figures

High-momentum two-particle correlations are a useful tool for studying jet-quenching effects in the quark-gluon plasma. Angular correlations between neutral-pion triggers and charged hadrons with transverse momenta in the range 4--12~GeV/$c$ and 0.5--7~GeV/$c$, respectively, have been measured by the PHENIX experiment in 2014 for Au$+$Au collisions at $\sqrt{s_{_{NN}}}=200$~GeV. Suppression is observed in the yield of high-momentum jet fragments opposite the trigger particle, which indicates jet suppression stemming from in-medium partonic energy loss, while enhancement is observed for low-momentum particles. The ratio and differences between the yield in Au$+$Au collisions and $p$$+$$p$ collisions, $I_{AA}$ and $\Delta_{AA}$, as a function of the trigger-hadron azimuthal separation, $\Delta\phi$, are measured for the first time at the Relativistic Heavy Ion Collider. These results better quantify how the yield of low-$p_T$ associated hadrons is enhanced at wide angle, which is crucial for studying energy loss as well as medium-response effects., Comment: 535 authors from 84 institutions, 12 pages, 8 figures. v2 is version accepted for publication in Physical Review C. HEPdata tables for the points plotted in figures for this and previous PHENIX publications are (or will be) publicly available at http://www.phenix.bnl.gov/papers.html

Classifier-free guidance (CFG) is a fundamental tool in modern diffusion models for text-guided generation. Although effective, CFG has notable drawbacks. For instance, DDIM with CFG lacks invertibility, complicating image editing; furthermore, high guidance scales, essential for high-quality outputs, frequently result in issues like mode collapse. Contrary to the widespread belief that these are inherent limitations of diffusion models, this paper reveals that the problems actually stem from the off-manifold phenomenon associated with CFG, rather than the diffusion models themselves. More specifically, inspired by the recent advancements of diffusion model-based inverse problem solvers (DIS), we reformulate text-guidance as an inverse problem with a text-conditioned score matching loss and develop CFG++, a novel approach that tackles the off-manifold challenges inherent in traditional CFG. CFG++ features a surprisingly simple fix to CFG, yet it offers significant improvements, including better sample quality for text-to-image generation, invertibility, smaller guidance scales, reduced mode collapse, etc. Furthermore, CFG++ enables seamless interpolation between unconditional and conditional sampling at lower guidance scales, consistently outperforming traditional CFG at all scales. Moreover, CFG++ can be easily integrated into high-order diffusion solvers and naturally extends to distilled diffusion models. Experimental results confirm that our method significantly enhances performance in text-to-image generation, DDIM inversion, editing, and solving inverse problems, suggesting a wide-ranging impact and potential applications in various fields that utilize text guidance. Project Page: https://cfgpp-diffusion.github.io/., Comment: 25 pages, 21 figures. Project Page: https://cfgpp-diffusion.github.io/

We present novel insights into the interplay between tidal forces and star formation in interacting galaxies before their first pericentre passage. We investigate seven close pair galaxies devoid of visible tidal disturbances, such as tails, bridges, and shells. Using integral field spectroscopy (IFS) data of extended Calar Alto Legacy Integral Field Area (eCALIFA), we unveil a previously unreported phenomenon: H alhpa emission, a proxy for recent star formation, exhibits a significant enhancement in regions facing the companion galaxy, reaching up to 1.9 times higher flux compared to opposite directions. Notably, fainter companions within pairs display a more pronounced one-sided H alpha excess, exceeding the typical range observed in isolated galaxies with 2 sigma confidence level. Furthermore, the observed H alpha excess in fainter companion galaxies exhibits a heightened prominence at the outer galactic regions. These findings suggest that tidal forces generated before the first pericentre passage exert a stronger influence on fainter galaxies due to their shallower potential wells by their brighter companions. This unveils a more intricate interplay between gravitational interactions and star formation history within interacting galaxies than previously understood, highlighting the need further to explore the early stages of interaction in galaxy evolution., Comment: 7 pages, 4 figgures, Accepted for publication in MNRAS Letters

Decoding methods for large language models (LLMs) usually struggle with the tradeoff between ensuring factuality and maintaining diversity. For example, a higher p threshold in the nucleus (top-p) sampling increases the diversity but decreases the factuality, and vice versa. In this paper, we propose REAL (Residual Entropy from Asymptotic Line) sampling, a decoding method that achieves improved factuality and diversity over nucleus sampling by predicting an adaptive threshold of $p$. Specifically, REAL sampling predicts the step-wise likelihood of an LLM to hallucinate, and lowers the p threshold when an LLM is likely to hallucinate. Otherwise, REAL sampling increases the p threshold to boost the diversity. To predict the step-wise hallucination likelihood without supervision, we construct a Token-level Hallucination Forecasting (THF) model to predict the asymptotic entropy (i.e., inherent uncertainty) of the next token by extrapolating the next-token entropies from a series of LLMs with different sizes. If a LLM's entropy is higher than the asymptotic entropy (i.e., the LLM is more uncertain than it should be), the THF model predicts a high hallucination hazard, which leads to a lower p threshold in REAL sampling. In the FactualityPrompts benchmark, we demonstrate that REAL sampling based on a 70M THF model can substantially improve the factuality and diversity of 7B LLMs simultaneously, judged by both retrieval-based metrics and human evaluation. After combined with contrastive decoding, REAL sampling outperforms 9 sampling methods, and generates texts that are more factual than the greedy sampling and more diverse than the nucleus sampling with $p=0.5$. Furthermore, the predicted asymptotic entropy is also a useful unsupervised signal for hallucination detection tasks.

SBND is the near detector of the Short-Baseline Neutrino program at Fermilab. Its location near to the Booster Neutrino Beam source and relatively large mass will allow the study of neutrino interactions on argon with unprecedented statistics. This paper describes the expected performance of the SBND photon detection system, using a simulated sample of beam neutrinos and cosmogenic particles. Its design is a dual readout concept combining a system of 120 photomultiplier tubes, used for triggering, with a system of 192 X-ARAPUCA devices, located behind the anode wire planes. Furthermore, covering the cathode plane with highly-reflective panels coated with a wavelength-shifting compound recovers part of the light emitted towards the cathode, where no optical detectors exist. We show how this new design provides a high light yield and a more uniform detection efficiency, an excellent timing resolution and an independent 3D-position reconstruction using only the scintillation light. Finally, the whole reconstruction chain is applied to recover the temporal structure of the beam spill, which is resolved with a resolution on the order of nanoseconds., Comment: 21 pages, 17 figures

The Carbon monOxide Mapping Array Project (COMAP) Pathfinder survey continues to demonstrate the feasibility of line-intensity mapping using high-redshift carbon monoxide (CO) line emission traced at cosmological scales. The latest COMAP Pathfinder power spectrum analysis is based on observations through the end of Season 2, covering the first three years of Pathfinder operations. We use our latest constraints on the CO(1-0) line-intensity power spectrum at $z\sim3$ to update corresponding constraints on the cosmological clustering of CO line emission and thus the cosmic molecular gas content at a key epoch of galaxy assembly. We first mirror the COMAP Early Science interpretation, considering how Season 2 results translate to limits on the shot noise power of CO fluctuations and the bias of CO emission as a tracer of the underlying dark matter distribution. The COMAP Season 2 results place the most stringent limits on the CO tracer bias to date, at $\langle{Tb}\rangle<4.8$ $\mu$K. These limits narrow the model space significantly compared to previous CO line-intensity mapping results while maintaining consistency with small-volume interferometric surveys of resolved line candidates. The results also express a weak preference for CO emission models used to guide fiducial forecasts from COMAP Early Science, including our data-driven priors. We also consider directly constraining a model of the halo-CO connection, and show qualitative hints of capturing the total contribution of faint CO emitters through the improved sensitivity of COMAP data. With continued observations and matching improvements in analysis, the COMAP Pathfinder remains on track for a detection of cosmological clustering of CO emission., Comment: 9 pages + bibliography and appendices (13 pages total); 9 figures, 1 table; v2 reflects minor changes made for version submitted to A&A, with no changes to top-line results

We present updated constraints on the cosmological 3D power spectrum of carbon monoxide CO(1-0) emission in the redshift range $2.4$-$3.4$. The constraints are derived from the two first seasons of Carbon monOxide Mapping Array Project (COMAP) Pathfinder line-intensity mapping observations aiming to trace star-formation during the Epoch of Galaxy Assembly. These results improve on the previous Early Science (ES) results through both increased data volume and improved data processing methodology. On the methodological side, we now perform cross-correlations between groups of detectors (''feed-groups''), as opposed to cross-correlations between single feeds, and this new feed-group pseudo power spectrum (FGPXS) is constructed to be more robust against systematic effects. In terms of data volume, the effective mapping speed is significantly increased due to an improved observational strategy as well as better data selection methodology. The updated spherically- and field-averaged FGPXS, $\tilde{C}(k)$, is consistent with zero, at a probability-to-exceed of around $34\,\%$, with an excess of $2.7\,\sigma$ in the most sensitive bin. Our power spectrum estimate is about an order of magnitude more sensitive in our six deepest bins across ${0.09\,\mathrm{Mpc}^{-1} < k < 0.73\,\mathrm{Mpc}^{-1}}$, as compared to the feed-feed pseudo power spectrum (FPXS) of COMAP ES. Each of these bins individually constrains the CO power spectrum to ${kP_\mathrm{CO}(k)< 2400-4900\,\mathrm{\mu K^2 Mpc^{2}}}$ at $95\,\%$ confidence. To monitor potential contamination from residual systematic effects, we analyze a set of 312 difference-map null tests and find that these are consistent with the instrumental noise prediction. In sum, these results provide the strongest direct constraints on the cosmological 3D CO(1-0) power spectrum published to date., Comment: 17 pages, 10 figures, v2 reflects changes made for version submitted to Astronomy and Astrophysics, addition of additional figure clarifying methodology, no change to final results

The CO Mapping Array Project (COMAP) Pathfinder is performing line intensity mapping of CO emission to trace the distribution of unresolved galaxies at redshift $z \sim 3$. We present an improved version of the COMAP data processing pipeline and apply this to the first two seasons of observations. This analysis improves on the COMAP Early Science (ES) results in several key aspects. On the observational side, all second season scans were made in constant-elevation mode, after noting that the previous Lissajous scans were associated with increased systematic errors; those scans accounted for 50% of the total Season 1 data volume. Secondly, all new observations were restricted to an elevation range of 35-65 degrees, to minimize sidelobe ground pickup. On the data processing side, more effective data cleaning in both the time- and map-domain has allowed us to eliminate all data-driven power spectrum-based cuts. This increases the overall data retention and reduces the risk of signal subtraction bias. On the other hand, due to the increased sensitivity, two new pointing-correlated systematic errors have emerged, and we introduce a new map-domain PCA filter to suppress these. Subtracting only 5 out of 256 PCA modes, we find that the standard deviation of the cleaned maps decreases by 67% on large angular scales, and after applying this filter, the maps appear consistent with instrumental noise. Combining all these improvements, we find that each hour of raw Season 2 observations yields on average 3.2 times more cleaned data compared to ES analysis. Combining this with the increase in raw observational hours, the effective amount of data available for high-level analysis is a factor of 8 higher than in ES. The resulting maps have reached an uncertainty of $25$-$50\,\mu K$ per voxel, providing by far the strongest constraints on cosmological CO line emission published to date., Comment: 23 pages, 22 figures, for submission to Astronomy and Astrophysics

Neuronal networking supports complex brain functions, with neurotransmitters facilitating communication through chemical synapses. The release probability of neurotransmitters varies and is influenced by pre-synaptic neuronal activity. Recent findings suggest that blocking astrocytic N-Methyl-D-Aspartate (NMDA) receptors reduces this variation. However, the theoretical implications of this reduction on neuronal dynamics have not been thoroughly investigated. Utilizing continuous attractor neural network (CANN) models with short-term synaptic depression (STD), we explore the effects of reduced release probability variation. Our results show that blocking astrocytic NMDA receptors stabilizes attractor states and diminishes their mobility. These insights enhance our understanding of NMDA receptors' role in astrocytes and their broader impact on neural computation and memory, with potential implications for neurological conditions involving NMDA receptor antagonists., Comment: 22 pages, 6 figures

As the scale of generative models continues to grow, efficient reuse and adaptation of pre-trained models have become crucial considerations. In this work, we propose Voicebox Adapter, a novel approach that integrates fine-grained conditions into a pre-trained Voicebox speech generation model using a cross-attention module. To ensure a smooth integration of newly added modules with pre-trained ones, we explore various efficient fine-tuning approaches. Our experiment shows that the LoRA with bias-tuning configuration yields the best performance, enhancing controllability without compromising speech quality. Across three fine-grained conditional generation tasks, we demonstrate the effectiveness and resource efficiency of Voicebox Adapter. Follow-up experiments further highlight the robustness of Voicebox Adapter across diverse data setups., Comment: Accepted by InterSpeech 2024

Self-driving vehicles have the potential to reduce accidents and fatalities on the road. Many production vehicles already come equipped with basic self-driving capabilities, but have trouble following lanes in adverse lighting and weather conditions. Therefore, we develop, analyze, and evaluate two vehicular lane-keeping algorithms under dynamic weather conditions using a combined deep learning- and hand-crafted approach and an end-to-end deep learning approach. We use image segmentation- and linear-regression based deep learning to drive the vehicle toward the center of the lane, measuring the amount of laps completed, average speed, and average steering error per lap. Our hybrid model completes more laps than our end-to-end deep learning model. In the future, we are interested in combining our algorithms to form one cohesive approach to lane-following., Comment: Supported by the National Science Foundation under Grants No. 2150292 and 2150096

We consider the two-sample testing problem for networks, where the goal is to determine whether two sets of networks originated from the same stochastic model. Assuming no vertex correspondence and allowing for different numbers of nodes, we address a fundamental network testing problem that goes beyond simple adjacency matrix comparisons. We adopt the stochastic block model (SBM) for network distributions, due to their interpretability and the potential to approximate more general models. The lack of meaningful node labels and vertex correspondence translate to a graph matching challenge when developing a test for SBMs. We introduce an efficient algorithm to match estimated network parameters, allowing us to properly combine and contrast information within and across samples, leading to a powerful test. We show that the matching algorithm, and the overall test are consistent, under mild conditions on the sparsity of the networks and the sample sizes, and derive a chi-squared asymptotic null distribution for the test. Through a mixture of theoretical insights and empirical validations, including experiments with both synthetic and real-world data, this study advances robust statistical inference for complex network data.

Motivated by the works of Erd\"os, Pomerance, Wolke and Harman on the sum-of-divisor function $\sigma(n)$, we study the distribution of a special class of natural numbers closely related to (multiply) perfect numbers which we term `$(\ell;k)$-within-perfect numbers', where $\ell >1$ is a real number and $k: [1, \infty) \rightarrow (0, \infty)$ is an increasing and unbounded function., Comment: Minor changes, Accepted for publication. This article together with arXiv:1610.04253v4 (published in Acta Arithmetica 194 (2020), 341-366) supersede arXiv:1610.04253v2

Image diffusion distillation achieves high-fidelity generation with very few sampling steps. However, applying these techniques directly to video diffusion often results in unsatisfactory frame quality due to the limited visual quality in public video datasets. This affects the performance of both teacher and student video diffusion models. Our study aims to improve video diffusion distillation while improving frame appearance using abundant high-quality image data. We propose motion consistency model (MCM), a single-stage video diffusion distillation method that disentangles motion and appearance learning. Specifically, MCM includes a video consistency model that distills motion from the video teacher model, and an image discriminator that enhances frame appearance to match high-quality image data. This combination presents two challenges: (1) conflicting frame learning objectives, as video distillation learns from low-quality video frames while the image discriminator targets high-quality images; and (2) training-inference discrepancies due to the differing quality of video samples used during training and inference. To address these challenges, we introduce disentangled motion distillation and mixed trajectory distillation. The former applies the distillation objective solely to the motion representation, while the latter mitigates training-inference discrepancies by mixing distillation trajectories from both the low- and high-quality video domains. Extensive experiments show that our MCM achieves the state-of-the-art video diffusion distillation performance. Additionally, our method can enhance frame quality in video diffusion models, producing frames with high aesthetic scores or specific styles without corresponding video data., Comment: NeurIPS 2024; project page: https://yhzhai.github.io/mcm/

We provide an alternative characterization of two-dimensional locality (necessary e.g. to define the Hall conductivity of a Fermi projection) using the spectral projections of the Laughlin flux operator. Using this abstract characterization, we define generalizations of this locality, which we term quasi-2D. We go on to calculate the path-connected components of spaces of unitaries or orthogonal projections which are quasi-2D-local and find a starkly different behavior compared with the actual 2D column of the Kitaev table, exhibiting e.g., in the unitary chiral case, infinitely many $\mathbb{Z}$-valued indices., Comment: 21 pages, 5 figures

The current automatic speaker verification (ASV) task involves making binary decisions on two types of trials: target and non-target. However, emerging advancements in speech generation technology pose significant threats to the reliability of ASV systems. This study investigates whether ASV effortlessly acquires robustness against spoofing attacks (i.e., zero-shot capability) by systematically exploring diverse ASV systems and spoofing attacks, ranging from traditional to cutting-edge techniques. Through extensive analyses conducted on eight distinct ASV systems and 29 spoofing attack systems, we demonstrate that the evolution of ASV inherently incorporates defense mechanisms against spoofing attacks. Nevertheless, our findings also underscore that the advancement of spoofing attacks far outpaces that of ASV systems, hence necessitating further research on spoofing-robust ASV methodologies., Comment: 5 pages, 3 figures, 3 tables, Interspeech 2024

To aid in prediction of turbulent boundary layer flows over rough surfaces, a new model is proposed to estimate hydrodynamic roughness based solely on geometric surface information. The model is based on a fluid-mechanics motivated geometric parameter called the wind-shade factor. Sheltering is included using a rapid algorithm adapted from the landscape shadow literature, while local pressure drag is estimated using a piecewise potential flow approximation. Similarly to evaluating traditional surface parameters such as skewness or average slope magnitude, the wind-shade factor is purely geometric and can be evaluated efficiently from knowing the surface elevation map and the mean flow direction. The wind-shade roughness model is applied to over 100 different surfaces available in a public roughness database and some others, and the predicted sandgrain-roughness heights are compared to measured values. Effects of various model ingredients are analyzed, and transitionally rough surfaces are treated by adding a term representing the viscous stress component., Comment: Paper submitted to J. Fluid Mechanics for publication, currently under review. Contains Notebooks but links will work only once it is published

Time series self-supervised learning (SSL) aims to exploit unlabeled data for pre-training to mitigate the reliance on labels. Despite the great success in recent years, there is limited discussion on the potential noise in the time series, which can severely impair the performance of existing SSL methods. To mitigate the noise, the de facto strategy is to apply conventional denoising methods before model training. However, this pre-processing approach may not fully eliminate the effect of noise in SSL for two reasons: (i) the diverse types of noise in time series make it difficult to automatically determine suitable denoising methods; (ii) noise can be amplified after mapping raw data into latent space. In this paper, we propose denoising-aware contrastive learning (DECL), which uses contrastive learning objectives to mitigate the noise in the representation and automatically selects suitable denoising methods for every sample. Extensive experiments on various datasets verify the effectiveness of our method. The code is open-sourced., Comment: Accepted to 33rd International Joint Conference on Artificial Intelligence (IJCAI 2024)

In light of growing threats posed by climate change in general and sea level rise (SLR) in particular, the necessity for computationally efficient means to estimate and analyze potential coastal flood hazards has become increasingly pressing. Data-driven supervised learning methods serve as promising candidates that can dramatically expedite the process, thereby eliminating the computational bottleneck associated with traditional physics-based hydrodynamic simulators. Yet, the development of accurate and reliable coastal flood prediction models, especially those based on Deep Learning (DL) techniques, has been plagued with two major issues: (1) the scarcity of training data and (2) the high-dimensional output required for detailed inundation mapping. To remove this barrier, we present a systematic framework for training high-fidelity Deep Vision-based coastal flood prediction models in low-data settings. We test the proposed workflow on different existing vision models, including a fully transformer-based architecture and a Convolutional Neural Network (CNN) with additive attention gates. Additionally, we introduce a deep CNN architecture tailored specifically to the coastal flood prediction problem at hand. The model was designed with a particular focus on its compactness so as to cater to resource-constrained scenarios and accessibility aspects. The performance of the developed DL models is validated against commonly adopted geostatistical regression methods and traditional Machine Learning (ML) approaches, demonstrating substantial improvement in prediction quality. Lastly, we round up the contributions by providing a meticulously curated dataset of synthetic flood inundation maps of Abu Dhabi's coast produced with a physics-based hydrodynamic simulator, which can serve as a benchmark for evaluating future coastal flood prediction models.

It is known that minimal Lagrangians in K\"ahler--Einstein manifolds of non-positive scalar curvature are linearly stable under Hamiltonian deformations. We prove that they are also stable under the Lagrangian mean curvature flow, and therefore establish the equivalence between linear stability and dynamical stability. Specifically, if one starts the mean curvature flow with a Lagrangian which is $C^1$-close and Hamiltonian isotopic to a minimal Lagrangian, the flow exists smoothly for all time, and converges to that minimal Lagrangian. Due to the work of Neves [Ann. of Math. 2013], this cannot be true for $C^0$-closeness., Comment: 21 pages

Supermassive black hole binaries (SMBHBs) present us with exciting opportunities for multi-messenger science. These systems are thought to form naturally in galaxy mergers and therefore have the potential to produce electromagnetic (EM) radiation as well as gravitational waves (GWs) detectable with pulsar timing arrays (PTAs). Once GWs from individually resolved SMBHBs are detected, the identification of the host galaxy will be a major challenge due to the ambiguity in possible EM signatures and the poor localization capability of PTAs. In order to aid EM observations in choosing which sources to follow up, we attempt to quantify the number of plausible hosts in both realistic and idealistic scenarios. We outline a host galaxy identification pipeline that injects a single-source GW signal into a simulated PTA dataset, uses production-level techniques to recover the signal, quantifies the localization region and number of galaxies contained therein, and finally imposes cuts on the galaxies using the binary parameters estimated from the GW search. In an ideal case, we find that the 90% credible areas span 29 deg^2 to 241 deg^2, containing about 14 to 341 galaxies. After cuts, the number of galaxies remaining ranges from 22 at worst to 1 (the true host) at best. In a more realistic case, if the signal is sufficiently localized, the sky areas range from 287 deg^2 to 530 deg^2 and enclose about 285 to 1238 galaxies. After cuts, the number of galaxies is 397 at worst and 27 at best. While the signal-to-noise ratio is the primary determinant of the localization area of a given source, we find that the size of the area is also influenced by the proximity of nearby pulsars on the sky and the chirp mass of the source., Comment: Submitted to ApJ

Accurate control of the total duration of generated speech by adjusting the speech rate is crucial for various text-to-speech (TTS) applications. However, the impact of adjusting the speech rate on speech quality, such as intelligibility and speaker characteristics, has been underexplored. In this work, we propose a novel total-duration-aware (TDA) duration model for TTS, where phoneme durations are predicted not only from the text input but also from an additional input of the total target duration. We also propose a MaskGIT-based duration model that enhances the diversity and quality of the predicted phoneme durations. Our results demonstrate that the proposed TDA duration models achieve better intelligibility and speaker similarity for various speech rate configurations compared to the baseline models. We also show that the proposed MaskGIT-based model can generate phoneme durations with higher quality and diversity compared to its regression or flow-matching counterparts., Comment: Accepted to Interspeech 2024

Transformers have rapidly become the preferred choice for audio classification, surpassing methods based on CNNs. However, Audio Spectrogram Transformers (ASTs) exhibit quadratic scaling due to self-attention. The removal of this quadratic self-attention cost presents an appealing direction. Recently, state space models (SSMs), such as Mamba, have demonstrated potential in language and vision tasks in this regard. In this study, we explore whether reliance on self-attention is necessary for audio classification tasks. By introducing Audio Mamba (AuM), the first self-attention-free, purely SSM-based model for audio classification, we aim to address this question. We evaluate AuM on various audio datasets - comprising six different benchmarks - where it achieves comparable or better performance compared to well-established AST model., Comment: Code is available at https://github.com/mhamzaerol/Audio-Mamba-AuM

Data subset selection aims to find a smaller yet informative subset of a large dataset that can approximate the full-dataset training, addressing challenges associated with training neural networks on large-scale datasets. However, existing methods tend to specialize in either high or low selection ratio regimes, lacking a universal approach that consistently achieves competitive performance across a broad range of selection ratios. We introduce a universal and efficient data subset selection method, Best Window Selection (BWS), by proposing a method to choose the best window subset from samples ordered based on their difficulty scores. This approach offers flexibility by allowing the choice of window intervals that span from easy to difficult samples. Furthermore, we provide an efficient mechanism for selecting the best window subset by evaluating its quality using kernel ridge regression. Our experimental results demonstrate the superior performance of BWS compared to other baselines across a broad range of selection ratios over datasets, including CIFAR-10/100 and ImageNet, and the scenarios involving training from random initialization or fine-tuning of pre-trained models., Comment: ICML 2024

This work presents an Iterative Constraint Energy Minimizing Generalized Multiscale Finite Element Method (ICEM-GMsFEM) for solving the contact problem with high contrast coefficients. The model problem can be characterized by a variational inequality, where we add a penalty term to convert this problem into a non-smooth and non-linear unconstrained minimizing problem. The characterization of the minimizer satisfies the variational form of a mixed Dirilect-Neumann-Robin boundary value problem. So we apply CEM-GMsFEM iteratively and introduce special boundary correctors along with multiscale spaces to achieve an optimal convergence rate. Numerical results are conducted for different highly heterogeneous permeability fields, validating the fast convergence of the CEM-GMsFEM iteration in handling the contact boundary and illustrating the stability of the proposed method with different sets of parameters. We also prove the fast convergence of the proposed iterative CEM-GMsFEM method and provide an error estimate of the multiscale solution under a mild assumption.

The adoption of Building Information Modeling (BIM) is beneficial in construction projects. However, it faces challenges due to the lack of a unified and scalable framework for converting 3D model details into BIM. This paper introduces SRBIM, a unified semantic reconstruction architecture for BIM generation. Our approach's effectiveness is demonstrated through extensive qualitative and quantitative evaluations, establishing a new paradigm for automated BIM modeling., Comment: CVPRW 2024, Oral

Precise segmentation of architectural structures provides detailed information about various building components, enhancing our understanding and interaction with our built environment. Nevertheless, existing outdoor 3D point cloud datasets have limited and detailed annotations on architectural exteriors due to privacy concerns and the expensive costs of data acquisition and annotation. To overcome this shortfall, this paper introduces a semantically-enriched, photo-realistic 3D architectural models dataset and benchmark for semantic segmentation. It features 4 different building purposes of real-world buildings as well as an open architectural landscape in Hong Kong. Each point cloud is annotated into one of 14 semantic classes., Comment: CVPRW 2024 (Oral)

We examined archival Far Ultraviolet Spectroscopic Explorer data to search for far-ultraviolet emission lines in the starburst galaxy M82. The observations were made in an outflow region that extends beyond the galactic disk. We found the O VI $\lambda\lambda$ 1032, 1038 emission lines from the galaxy's southern outflow region. The O VI lines suggest that the outflowing warm-hot gas is undergoing radiative cooling. We measured a radial velocity of $\sim$420 km s$^{-1}$ from the O VI lines, which is faster than the velocity seen in H$\alpha$ observations. The O VI $\lambda$1038 emission line seems to be blended with the C II $\lambda$1037 emission line, which has a radial velocity of $\sim$300 km s$^{-1}$, similar to what is observed in H$\alpha$ observations. The outflow medium of M82 appears to be composed of gas in multiple phases with varying temperatures and kinematics. Future spectroscopic observations in high energy regimes covering a wider spatial area are necessary to understand better the properties of the warm-hot gas medium in the outflow., Comment: 9 pages, 5 figures, Accepted to AJ

The prevailing approaches in Network Intrusion Detection Systems (NIDS) are often hampered by issues such as high resource consumption, significant computational demands, and poor interpretability. Furthermore, these systems generally struggle to identify novel, rapidly changing cyber threats. This paper delves into the potential of incorporating Neurosymbolic Artificial Intelligence (NSAI) into NIDS, combining deep learning's data-driven strengths with symbolic AI's logical reasoning to tackle the dynamic challenges in cybersecurity, which also includes detailed NSAI techniques introduction for cyber professionals to explore the potential strengths of NSAI in NIDS. The inclusion of NSAI in NIDS marks potential advancements in both the detection and interpretation of intricate network threats, benefiting from the robust pattern recognition of neural networks and the interpretive prowess of symbolic reasoning. By analyzing network traffic data types and machine learning architectures, we illustrate NSAI's distinctive capability to offer more profound insights into network behavior, thereby improving both detection performance and the adaptability of the system. This merging of technologies not only enhances the functionality of traditional NIDS but also sets the stage for future developments in building more resilient, interpretable, and dynamic defense mechanisms against advanced cyber threats. The continued progress in this area is poised to transform NIDS into a system that is both responsive to known threats and anticipatory of emerging, unseen ones.

We report a free-floating planet (FFP) candidate identified from the analysis of the microlensing event KMT-2023-BLG-2669. The lensing light curve is characterized by a short duration $(\lesssim 3\,{\rm days})$ and a small amplitude $(\lesssim 0.7\,{\rm mag})$. From the analysis, we find the Einstein timescale of $t_{\rm E} \backsimeq 0.33\,{\rm days}$ and the Einstein radius of $\theta_{\rm E} \backsimeq 4.41\,{\mu}{\rm as}$. These measurements enable us to infer the lens mass as $M = 8\,M_{\oplus} (\pi_{\rm rel} / 0.1\,{\rm mas})^{-1}$, where $\pi_{\rm rel}$ is the relative lens-source parallax. The inference implies that the lens is a sub-Neptune- to Saturn-mass object depending on its unknown distance. This is the ninth isolated planetary-mass microlens with $\theta_{\rm E} < 10\,{\mu}{\rm as}$, which (as shown by \citealt{gould22}) is a useful threshold for a FFP candidate. We conduct extensive searches for possible signals of a host star in the light curve, but find no strong evidence for the host. We investigate the possibility of using late-time high-resolution imaging to probe for possible hosts. In particular, we discuss that for the case of finite-source point-lens FFP candidates, it would be possible to search for very wide separation hosts immediately, although such searches are "high-risk, high-reward"., Comment: 26 pages, 6 figures, accepted in publication in AJ

We consider the solution of nonlinear inverse problems where the forward problem is a discretization of a partial differential equation. Such problems are notoriously difficult to solve in practice and require minimizing a combination of a data-fit term and a regularization term. The main computational bottleneck of typical algorithms is the direct estimation of the data misfit. Therefore, likelihood-free approaches have become appealing alternatives. Nonetheless, difficulties in generalization and limitations in accuracy have hindered their broader utility and applicability. In this work, we use a paired autoencoder framework as a likelihood-free estimator for inverse problems. We show that the use of such an architecture allows us to construct a solution efficiently and to overcome some known open problems when using likelihood-free estimators. In particular, our framework can assess the quality of the solution and improve on it if needed. We demonstrate the viability of our approach using examples from full waveform inversion and inverse electromagnetic imaging., Comment: 18 pages, 6 figures

We investigate the nature of the short-term anomaly that appears in the lensing light curve of KMT-2023-BLG-1866. The anomaly was only partly covered due to its short duration, less than a day, coupled with cloudy weather conditions and restricted nighttime duration. Considering intricacy of interpreting partially covered signals, we thoroughly explore all potential degenerate solutions. Through this process, we identify three planetary scenarios that equally well account for the observed anomaly. These scenarios are characterized by the specific planetary parameters: $(s, q)_{\rm inner} = [0.9740 \pm 0.0083, (2.46 \pm 1.07) \times 10^{-5}]$, $(s, q)_{\rm intermediate} = [0.9779 \pm 0.0017, (1.56 \pm 0.25)\times 10^{-5}]$, and $(s, q)_{\rm outer} = [0.9894 \pm 0.0107, (2.31 \pm 1.29)\times 10^{-5}]$, where $s$ and $q$ denote the projected separation (scaled to the Einstein radius) and mass ratio between the planet and its host, respectively. We identify that the ambiguity between the inner and outer solutions stems from the inner-outer degeneracy, while the similarity between the intermediate solution and the others is due to an accidental degeneracy caused by incomplete anomaly coverage. Through Bayesian analysis utilizing the constraints derived from measured lensing observables and blending flux, our estimation indicates that the lens system comprises a very low-mass planet orbiting an early M-type star situated approximately (6.2 -- 6.5)~kpc from Earth in terms of median posterior values for the different solutions. The median mass of the planet host is in the range of (0.48 -- 0.51)~$M_\odot$, and that of the planet's mass spans a range of (2.6 -- 4.0)~$M_{\rm E}$, varying across different solutions. The detection of KMT-2023-BLG-1866Lb signifies the extension of the lensing surveys to very low-mass planets that have been difficult to be detected from earlier surveys., Comment: 9 pages, 8 figures, 4 tables

A cautious interpretation of AI regulations and policy in the EU and the USA place explainability as a central deliverable of compliant AI systems. However, from a technical perspective, explainable AI (XAI) remains an elusive and complex target where even state of the art methods often reach erroneous, misleading, and incomplete explanations. "Explainability" has multiple meanings which are often used interchangeably, and there are an even greater number of XAI methods - none of which presents a clear edge. Indeed, there are multiple failure modes for each XAI method, which require application-specific development and continuous evaluation. In this paper, we analyze legislative and policy developments in the United States and the European Union, such as the Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, the AI Act, the AI Liability Directive, and the General Data Protection Regulation (GDPR) from a right to explanation perspective. We argue that these AI regulations and current market conditions threaten effective AI governance and safety because the objective of trustworthy, accountable, and transparent AI is intrinsically linked to the questionable ability of AI operators to provide meaningful explanations. Unless governments explicitly tackle the issue of explainability through clear legislative and policy statements that take into account technical realities, AI governance risks becoming a vacuous "box-ticking" exercise where scientific standards are replaced with legalistic thresholds, providing only a false sense of security in XAI., Comment: AI Governance Workshop at the 2024 International Joint Conference on Artificial Intelligence (IJCAI)

Latency attacks against object detection represent a variant of adversarial attacks that aim to inflate the inference time by generating additional ghost objects in a target image. However, generating ghost objects in the black-box scenario remains a challenge since information about these unqualified objects remains opaque. In this study, we demonstrate the feasibility of generating ghost objects in adversarial examples by extending the concept of "steal now, decrypt later" attacks. These adversarial examples, once produced, can be employed to exploit potential vulnerabilities in the AI service, giving rise to significant security concerns. The experimental results demonstrate that the proposed attack achieves successful attacks across various commonly used models and Google Vision API without any prior knowledge about the target model. Additionally, the average cost of each attack is less than \$ 1 dollars, posing a significant threat to AI security.

We undertake a project to reexamine microlensing data gathered from high-cadence surveys. The aim of the project is to reinvestigate lensing events with light curves exhibiting intricate anomaly features associated with caustics, yet lacking prior proposed models to explain these features. Through detailed reanalyses considering higher-order effects, we identify that accounting for orbital motions of lenses is vital in accurately explaining the anomaly features observed in the light curves of the lensing events OGLE-2018-BLG-0971, MOA-2023-BLG-065, and OGLE-2023-BLG-0136. We estimate the masses and distances to the lenses by conducting Bayesian analyses using the lensing parameters of the newly found lensing solutions. From these analyses, we identify that the lenses of the events OGLE-2018-BLG-0971 and MOA-2023-BLG-065 are binaries composed of M dwarfs, while the lens of OGLE-2023-BLG-0136 is likely to be a binary composed of an early K-dwarf primary and a late M-dwarf companion. For all lensing events, the probability of the lens residing in the bulge is considerably higher than that of it being located in the disk., Comment: 11 pages, 13 figures, 6 tables

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.