Search

For each punctured curve over a finite field, we construct local systems which do not come from a family of abelian varieties. We do so by proving a criterion which must be satisfied by local systems which do come from abelian varieties, inspired by an analogous Hodge theoretic criterion in characteristic zero. Our tools include $F$-isocrystals and some $p$-adic Hodge theory.

We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development.

The process of human speech production involves coordinated respiratory action to elicit acoustic speech signals. Typically, speech is produced when air is forced from the lungs and is modulated by the vocal tract, where such actions are interspersed by moments of breathing in air (inhalation) to refill the lungs again. Respiratory rate (RR) is a vital metric that is used to assess the overall health, fitness, and general well-being of an individual. Existing approaches to measure RR (number of breaths one takes in a minute) are performed using specialized equipment or training. Studies have demonstrated that machine learning algorithms can be used to estimate RR using bio-sensor signals as input. Speech-based estimation of RR can offer an effective approach to measure the vital metric without requiring any specialized equipment or sensors. This work investigates a machine learning based approach to estimate RR from speech segments obtained from subjects speaking to a close-talking microphone device. Data were collected from N=26 individuals, where the groundtruth RR was obtained through commercial grade chest-belts and then manually corrected for any errors. A convolutional long-short term memory network (Conv-LSTM) is proposed to estimate respiration time-series data from the speech signal. We demonstrate that the use of pre-trained representations obtained from a foundation model, such as Wav2Vec2, can be used to estimate respiration-time-series with low root-mean-squared error and high correlation coefficient, when compared with the baseline. The model-driven time series can be used to estimate $RR$ with a low mean absolute error (MAE) ~ 1.6 breaths/min., Comment: 8 pages, 6 figures, BioKDD workshop paper

With 95% of Internet traffic now encrypted, an effective approach to classifying this traffic is crucial for network security and management. This paper introduces ECHO -- a novel optimization process for ML/DL-based encrypted traffic classification. ECHO targets both classification time and memory utilization and incorporates two innovative techniques. The first component, HO (Hyperparameter Optimization of binnings), aims at creating efficient traffic representations. While previous research often uses representations that map packet sizes and packet arrival times to fixed-sized bins, we show that non-uniform binnings are significantly more efficient. These non-uniform binnings are derived by employing a hyperparameter optimization algorithm in the training stage. HO significantly improves accuracy given a required representation size, or, equivalently, achieves comparable accuracy using smaller representations. Then, we introduce EC (Early Classification of traffic), which enables faster classification using a cascade of classifiers adapted for different exit times, where classification is based on the level of confidence. EC reduces the average classification latency by up to 90\%. Remarkably, this method not only maintains classification accuracy but also, in certain cases, improves it. Using three publicly available datasets, we demonstrate that the combined method, Early Classification with Hyperparameter Optimization (ECHO), leads to a significant improvement in classification efficiency.

This paper addresses the problem of accurately estimating a function on one domain when only its discrete samples are available on another domain. To answer this challenge, we utilize a neural network, which we train to incorporate prior knowledge of the function. In addition, by carefully analyzing the problem, we obtain a bound on the error over the extrapolation domain and define a condition number for this problem that quantifies the level of difficulty of the setup. Compared to other machine learning methods that provide time series prediction, such as transformers, our approach is suitable for setups where the interpolation and extrapolation regions are general subdomains and, in particular, manifolds. In addition, our construction leads to an improved loss function that helps us boost the accuracy and robustness of our neural network. We conduct comprehensive numerical tests and comparisons of our extrapolation versus standard methods. The results illustrate the effectiveness of our approach in various scenarios., Comment: 32 pages, 11 figures

Eye-tracking technology is integral to numerous consumer electronics applications, particularly in the realm of virtual and augmented reality (VR/AR). These applications demand solutions that excel in three crucial aspects: low-latency, low-power consumption, and precision. Yet, achieving optimal performance across all these fronts presents a formidable challenge, necessitating a balance between sophisticated algorithms and efficient backend hardware implementations. In this study, we tackle this challenge through a synergistic software/hardware co-design of the system with an event camera. Leveraging the inherent sparsity of event-based input data, we integrate a novel sparse FPGA dataflow accelerator customized for submanifold sparse convolution neural networks (SCNN). The SCNN implemented on the accelerator can efficiently extract the embedding feature vector from each representation of event slices by only processing the non-zero activations. Subsequently, these vectors undergo further processing by a gated recurrent unit (GRU) and a fully connected layer on the host CPU to generate the eye centers. Deployment and evaluation of our system reveal outstanding performance metrics. On the Event-based Eye-Tracking-AIS2024 dataset, our system achieves 81% p5 accuracy, 99.5% p10 accuracy, and 3.71 Mean Euclidean Distance with 0.7 ms latency while only consuming 2.29 mJ per inference. Notably, our solution opens up opportunities for future eye-tracking systems. Code is available at https://github.com/CASR-HKU/ESDA/tree/eye_tracking., Comment: Accepted to CVPR 2024 workshop, AIS: Vision, Graphics, and AI for Streaming

This survey reviews the AIS 2024 Event-Based Eye Tracking (EET) Challenge. The task of the challenge focuses on processing eye movement recorded with event cameras and predicting the pupil center of the eye. The challenge emphasizes efficient eye tracking with event cameras to achieve good task accuracy and efficiency trade-off. During the challenge period, 38 participants registered for the Kaggle competition, and 8 teams submitted a challenge factsheet. The novel and diverse methods from the submitted factsheets are reviewed and analyzed in this survey to advance future event-based eye tracking research., Comment: Qinyu Chen is the corresponding author

Let $U/K$ be a smooth affine curve over a number field and let $L$ be an irreducible rank 3 $\overline{\mathbb Q}_{\ell}$-local system on $U$ with trivial determinant and infinite geometric monodromy around a cusp. Suppose further that $L$ extends to an integral model such that the Frobenius traces are contained in a fixed totally real number field. Then, after potentially shrinking $U$, there exists an abelian scheme $f\colon B_U\rightarrow U$ such that $L$ is a summand of $R^2f_*\overline{\mathbb Q}_{\ell}(1)$. The key ingredients are: (1) the totally real assumption implies $L$ admits a square root $M$; (2) the trace field of $M$ is sufficiently bounded, allowing us to use recent work of Krishnamoorthy-Yang-Zuo to construct an abelian scheme over $U_{\bar K}$ geometrically realizing $L$; and (3) Deligne's weight-monodromy theorem and the Rapoport-Zink spectral sequence, which allow us to pin down the arithmetizations using the total degeneration., Comment: 3 pages, comments welcome!

Many programming frameworks have been introduced to support the development of differentially private software applications. In this chapter, we survey some of the conceptual ideas underlying these frameworks in a way that we hope will be helpful for both practitioners and researchers. For practitioners, the survey can provide a starting point for understanding what features may be valuable when selecting a programming framework. For researchers, it can help organize existing work in a unified way and provide context for understanding new features in future frameworks., Comment: To appear as a chapter in the book "Differential Privacy for Artificial Intelligence," edited by Ferdinando Fioretto and Pascal van Hentenryck and to be published by now publishers

Executing smart contracts is a compute and storage-intensive task, which currently dominates modern blockchain's performance. Given that computers are becoming increasingly multicore, concurrency is an attractive approach to improve programs' execution runtime. A unique challenge of blockchains is that all replicas (miners or validators) must execute all smart contracts in the same logical order to maintain the semantics of State Machine Replication (SMR). In this work, we study the maximal level of parallelism attainable when focusing on the conflict graph between transactions packaged in the same block. This exposes a performance vulnerability that block creators may exploit against existing blockchain concurrency solutions, which rely on a total ordering phase for maintaining consistency amongst all replicas. To facilitate the formal aspects of our study, we develop a novel generic framework for Active State Machine Replication (ASMR) that is strictly serializable. We introduce the concept of graph scheduling and the definition of the minimal latency scheduling problem, which we prove to be NP-hard. We show that the restricted version of this problem for homogeneous transactions is equivalent to the classic Graph Vertex Coloring Problem, yet show that the heterogeneous case is more complex. We discuss the practical implications of these results., Comment: Updated version for conference version; 53 pages, 12 figures, LaTeX with Auxiliary Files, short single line

In the pursuit of reducing the number of trainable parameters in deep transformer networks, we employ Reinforcement Learning to dynamically select layers during training and tie them together. Every few iterations, the RL agent is asked whether to train each layer $i$ independently or to copy the weights of a previous layer $j

Event-based vision represents a paradigm shift in how vision information is captured and processed. By only responding to dynamic intensity changes in the scene, event-based sensing produces far less data than conventional frame-based cameras, promising to springboard a new generation of high-speed, low-power machines for edge intelligence. However, processing such dynamically sparse input originated from event cameras efficiently in real time, particularly with complex deep neural networks (DNN), remains a formidable challenge. Existing solutions that employ GPUs and other frame-based DNN accelerators often struggle to efficiently process the dynamically sparse event data, missing the opportunities to improve processing efficiency with sparse data. To address this, we propose ESDA, a composable dynamic sparse dataflow architecture that allows customized DNN accelerators to be constructed rapidly on FPGAs for event-based vision tasks. ESDA is a modular system that is composed of a set of parametrizable modules for each network layer type. These modules share a uniform sparse token-feature interface and can be connected easily to compose an all-on-chip dataflow accelerator on FPGA for each network model. To fully exploit the intrinsic sparsity in event data, ESDA incorporates the use of submanifold sparse convolutions that largely enhance the activation sparsity throughout the layers while simplifying hardware implementation. Finally, a network architecture and hardware implementation co-optimizing framework that allows tradeoffs between accuracy and performance is also presented. Experimental results demonstrate that when compared with existing GPU and hardware-accelerated solutions, ESDA achieves substantial speedup and improvement in energy efficiency across different applications, and it allows much wider design space for real-world deployments., Comment: Accepted to FPGA'24

Suppose $Y$ is a smooth variety equipped with a top form. We prove a simple theorem giving a sharp lower bound on the geometric genus of a family of subvarieties of $Y$, in terms of the dimension of this family. Two elementary applications are presented. On the one hand, we show that for a very general curve $C$ and a very general hypersurface $Y\subset \mathbb P^{n+1}$ of degree $\ge 2n+1$, any map $C \to Y$ is constant. On the other hand, we give a lower bound on the genus of a family of curves with an isotrivial factor in the associated family of Jacobians; we also characterize the families of curves attaining this bound as the families of degree $2$ branched covers of a fixed curve., Comment: 14 pages, comments welcome

Two moons of Saturn, Janus and Epimetheus, are in co-orbital motion, exchanging orbits approximately every four Earth years as the inner moon approaches the outer moon and they gravitationally interact. The orbital radii of these moons differ by only 50 km (less than the moons' mean physical radii), and it is this slight difference in their orbits that enables their periodic exchanges. Numerical n-body simulations can accurately model these exchanges using only Newtonian physics acting upon three objects: Saturn, Janus, and Epimetheus. Here we present analytical approaches and solutions, and corresponding computer simulations, designed to explore the effects of the initial orbital radius difference on otherwise similar co-orbital systems. Comparison with our simulation results illustrates that our analytic expressions provide very accurate predictions for the moon separations at closest approach and simulated post-exchange orbital radii. Our analytic estimates of the exchange period also match the simulated value for Janus and Epimetheus to within a few percent, although systems with smaller differences in their orbital radii are less well-modeled by our simple approach, suggesting that either full simulations or more sophisticated analytic approaches would be required to estimate exchange periods in those cases., Comment: 15 pages, 12 figures, submitted to Celestial Mechanics and Dynamical Astronomy

High-dimensional imbalanced data poses a machine learning challenge. In the absence of sufficient or high-quality labels, unsupervised feature selection methods are crucial for the success of subsequent algorithms. Therefore, we introduce a Marginal Laplacian Score (MLS), a modification of the well known Laplacian Score (LS) tailored to better address imbalanced data. We introduce an assumption that the minority class or anomalous appear more frequently in the margin of the features. Consequently, MLS aims to preserve the local structure of the dataset's margin. We propose its integration into modern feature selection methods that utilize the Laplacian score. We integrate the MLS algorithm into the Differentiable Unsupervised Feature Selection (DUFS), resulting in DUFS-MLS. The proposed methods demonstrate robust and improved performance on synthetic and public datasets., Comment: 10 pages

The COVID-19 pandemic has highlighted the need to upgrade systems for infectious disease surveillance and forecasting and modeling of the spread of infection, both of which inform evidence-based public health guidance and policies. Here, we discuss requirements for an effective surveillance system to support decision making during a pandemic, drawing on the lessons of COVID-19 in the U.S., while looking to jurisdictions in the U.S. and beyond to learn lessons about the value of specific data types. In this report, we define the range of decisions for which surveillance data are required, the data elements needed to inform these decisions and to calibrate inputs and outputs of transmission-dynamic models, and the types of data needed to inform decisions by state, territorial, local, and tribal health authorities. We define actions needed to ensure that such data will be available and consider the contribution of such efforts to improving health equity.

We consider a self-supervised approach to anomaly detection in tabular data. Random transformations are applied to the data, and then each transformation is identified based on its output. These predicted transformations are used to identify anomalies. In tabular data this approach faces many challenges that are related to the uncorrelated nature of the data. These challenges affect the transformations that should be used, as well as the use of their predictions. To this end, we propose SORTAD, a novel algorithm that is tailor-made to solve these challenges. SORTAD optimally chooses random transformations that help the classification process, and have a scoring function that is more sensitive to the changes in the transformations classification prediction encountered in tabular data. SORTAD achieved state-of-the-art results on multiple commonly used anomaly detection data sets, as well as in the overall results across all data sets tested., Comment: 16 pages, 7 figures

Background: The mapping of tree species within Norwegian forests is a time-consuming process, involving forest associations relying on manual labeling by experts. The process can involve both aerial imagery, personal familiarity, or on-scene references, and remote sensing data. The state-of-the-art methods usually use high resolution aerial imagery with semantic segmentation methods. Methods: We present a deep learning based tree species classification model utilizing only lidar (Light Detection And Ranging) data. The lidar images are segmented into four classes (Norway Spruce, Scots Pine, Birch, background) with a U-Net based network. The model is trained with focal loss over partial weak labels. A major benefit of the approach is that both the lidar imagery and the base map for the labels have free and open access. Results: Our tree species classification model achieves a macro-averaged F1 score of 0.70 on an independent validation with National Forest Inventory (NFI) in-situ sample plots. That is close to, but below the performance of aerial, or aerial and lidar combined models.

Estimating building footprint maps from geospatial data is of paramount importance in urban planning, development, disaster management, and various other applications. Deep learning methodologies have gained prominence in building segmentation maps, offering the promise of precise footprint extraction without extensive post-processing. However, these methods face challenges in generalization and label efficiency, particularly in remote sensing, where obtaining accurate labels can be both expensive and time-consuming. To address these challenges, we propose terrain-aware self-supervised learning, tailored to remote sensing, using digital elevation models from LiDAR data. We propose to learn a model to differentiate between bare Earth and superimposed structures enabling the network to implicitly learn domain-relevant features without the need for extensive pixel-level annotations. We test the effectiveness of our approach by evaluating building segmentation performance on test datasets with varying label fractions. Remarkably, with only 1% of the labels (equivalent to 25 labeled examples), our method improves over ImageNet pre-training, showing the advantage of leveraging unlabeled data for feature extraction in the domain of remote sensing. The performance improvement is more pronounced in few-shot scenarios and gradually closes the gap with ImageNet pre-training as the label fraction increases. We test on a dataset characterized by substantial distribution shifts and labeling errors to demonstrate the generalizability of our approach. When compared to other baselines, including ImageNet pretraining and more complex architectures, our approach consistently performs better, demonstrating the efficiency and effectiveness of self-supervised terrain-aware feature learning.

We report on a highly selective experimental setup for particle-$\gamma$ coincidence experiments at the Super-Enge Split-Pole Spectrograph (SE-SPS) of the John D. Fox Superconducting Linear Accelerator Laboratory at Florida State University (FSU) using fast CeBr$_3$ scintillators for $\gamma$-ray detection. Specifically, we report on the results of characterization tests for the first five CeBr$_3$ scintillation detectors of the CeBr$_3$ Array (CeBrA) with respect to energy resolution and timing characteristics. We also present results from the first particle-$\gamma$ coincidence experiments successfully performed with the CeBrA demonstrator and the FSU SE-SPS. We show that with the new setup, $\gamma$-decay branching ratios and particle-$\gamma$ angular correlations can be measured very selectively using narrow excitation energy gates, which are possible thanks to the excellent particle energy resolution of the SE-SPS. In addition, we highlight that nuclear level lifetimes in the nanoseconds regime can be determined by measuring the time difference between particle detection with the SE-SPS focal-plane scintillator and $\gamma$-ray detection with the fast CeBrA detectors. Selective excitation energy gates with the SE-SPS exclude any feeding contributions to these lifetimes.

Background: Accurate chamber volumetry from gated, non-contrast cardiac CT (NCCT) scans can be useful for potential screening of heart failure. Objectives: To validate a new, fully automated, AI-based method for cardiac volume and myocardial mass quantification from NCCT scans compared to contrasted CT Angiography (CCTA). Methods: Of a retrospectively collected cohort of 1051 consecutive patients, 420 patients had both NCCT and CCTA scans at mid-diastolic phase, excluding patients with cardiac devices. Ground truth values were obtained from the CCTA scans. Results: The NCCT volume computation shows good agreement with ground truth values. Volume differences [95% CI ] and correlation coefficients were: -9.6 [-45; 26] mL, r = 0.98 for LV Total, -5.4 [-24; 13] mL, r = 0.95 for LA, -8.7 [-45; 28] mL, r = 0.94 for RV, -5.2 [-27; 17] mL, r = 0.92 for RA, -3.2 [-42; 36] mL, r = 0.91 for LV blood pool, and -6.7 [-39; 26] g, r = 0.94 for LV wall mass, respectively. Mean relative volume errors of less than 7% were obtained for all chambers. Conclusions: Fully automated assessment of chamber volumes from NCCT scans is feasible and correlates well with volumes obtained from contrast study., Comment: Full version of JCCT technical report. Journal of Cardiovascular Computed Tomography (2023)

We performed a measurement of the $^{52}$Cr$(d,p)^{53}$Cr reaction at 16 MeV using the Florida State University Super-Enge Split-Pole Spectrograph (SE-SPS) and observed 26 states. While all of the states observed here had been seen in previous $(d,p)$ experiments, we changed five $L$ assignments from those reported previously and determined $L$ values for nine states that had not had such assignments made previously. The $g_{9/2}$ neutron strength observed in $^{53}$Cr in the present work and in the $N=29$ isotones $^{49}$Ca, $^{51}$Ti, and $^{55}$Fe via $(d,p)$ reactions is much smaller than the sum rule for this strength. Most of the observed $L=4$ strength in these nuclei is located in states near 4 MeV excitation energy. The remaining $g_{9/2}$ strength may be located in the continuum or may be fragmented among many bound states. A covariant density functional theory calculation provides support for the hypothesis that the $g_{9/2}$ neutron orbit is unbound in $^{53}$Cr. The ($\alpha,^3$He) reaction may provide a more sensitive probe for the missing $g_{9/2}$ neutron strength. In addition, particle-$\gamma$ coincidence experiments may help resolve some remaining questions in this nucleus., Comment: 15 pages, 5 figures. arXiv admin note: text overlap with arXiv:2212.04384