arXiv subjects

Yinzhi Wang

Publications and source records attributed to Yinzhi Wang.

18 recordsLinked to original sources

arXiv2026

Are cluster automorphism groups finitely generated?

This paper investigates the finite generation of cluster automorphism groups. By applying the pseudo $\mathbb{N}$-grading introduced in our previous work, we establish a sufficient condition for a cluster automorphism group to be finitely generated. As applications, we re-establish the finite generation of the automorphism groups for all finite mutation type cluster algebras, and verify the acyclic cases. Furthermore, we illustrate through examples that our approach significantly simplifies the computation of presentations for these groups in certain cases.

math.RA↗

arXiv2026

EdgeCrafter: Compact ViTs for Edge Dense Prediction via Task-Specialized Distillation

Deploying high-performance dense prediction models on resource-constrained edge devices remains challenging due to strict computation and memory budgets. In practice, lightweight systems for object detection, instance segmentation, and pose estimation are still dominated by CNN-based architectures such as YOLO, while compact Vision Transformers (ViTs) often struggle to achieve comparable accuracy-efficiency trade-offs, even with large-scale pretraining. We argue that this gap arises primarily from insufficient task-specific representation learning in small-scale ViTs, rather than from an inherent mismatch between ViTs and edge dense prediction. To address this issue, we introduce EdgeCrafter, a unified compact ViT framework for edge dense prediction centered on ECDet, a detection model built on a distilled compact backbone and an edge-friendly encoder-decoder design. The resulting detection-distilled representation transfers directly to instance segmentation and human pose estimation through lightweight task-specific prediction modules. Without using task annotations beyond COCO, ECDet-S achieves 51.7 box AP with fewer than 10M parameters, while ECInsSeg-X and ECPose-X reach 48.4 mask AP and 74.8 keypoint AP, respectively. As a complementary but more compute-intensive setting, Objects365 detection pretraining consistently improves performance across all scales, with the X variants reaching 59.9 box AP, 49.8 mask AP, and 75.9 keypoint AP. These results show that compact ViTs, when combined with task-specialized distillation and edge-aware design, can be a practical and competitive solution for edge dense prediction. Code is available at: https://intellindust-ai-lab.github.io/projects/EdgeCrafter

cs.CV↗

arXiv2026

CoT-X: An Adaptive Framework for Cross-Model Chain-of-Thought Transfer and Optimization

Long Chain-of-Thought (CoT) traces can improve reasoning accuracy, but repeatedly generating them is costly for smaller or latency-constrained language models. This paper studies a practical alternative: produce a rich rationale once with a capable \emph{thinking} model, compress it, and reuse the compressed trace as context for a cheaper \emph{answering} model. We introduce CoT-X, an adaptive framework for cross-model CoT transfer. CoT-X segments reasoning traces into semantic units, scores their diagnostic and logical importance, selects budget-feasible evidence paths, and reconstructs a coherent compressed rationale for the answering model. On $7,501$ Japanese medical licensing questions spanning $10$ specialties, CoT-X improves accuracy over direct truncation by up to $40.5\%$ under the same token budget, with the largest gains at $64$--$256$ tokens. Across $64$ thinking--answering pairs from eight DeepSeek-R1 and Qwen3 models (1.5B--32B parameters), reasoning transfer is most reliable within a model family, yet remains effective across families once compression normalizes the trace. A Gaussian Process Bayesian optimization layer finds near-optimal model--budget configurations with $15$ evaluations rather than an exhaustive search over all $64$ pairs, reducing evaluation cost by $84\%$. These results show that reasoning quality, token budget, and model compatibility can be optimized jointly, making CoT-style reasoning more practical under realistic deployment constraints.

cs.AI↗

arXiv2026

Evaluating PhaseNet on Teleseismic Data with MsPASS

Numerous studies have shown that the machine-learning picker PhaseNet produces accurate P and S picks on local earthquake signals, but its performance can degrade sharply on teleseismic signals. To address this limitation, we present a reproducible MsPASS workflow that (i) enables scalable data preparation and management for large seismic archives and (ii) supports standardized PhaseNet training and inference. We assembled a control dataset of 1.6 million waveforms linked to teleseismic P-wave picks made by analysts at the USArray Array Network Facility (ANF). The control dataset confirms that the PhaseNet model trained on regional signals performs poorly on these data. We then trained PhaseNet from scratch on the training split of the ANF control dataset and evaluated it on a non-overlapping held-out test split, increasing P-pick recall by 741.5% and yielding 683.9% more picks within a 0.1s residual window. We also evaluated PhaseNet across different model sizes on both CPUs and GPUs. Increasing the model size by about 120 times improved precision and recall by 15.6% and 23.2%, respectively. However, the scaled model reduced inference throughput by 87.2% on an NVIDIA A100 GPU and by 97.3% on a 128-core high-performance CPU node. These results indicate that scaling PhaseNet is more practical on GPUs than on CPUs, and that simply enlarging the model is not an efficient way to achieve large accuracy gains.

physics.geo-ph↗

arXiv2026

A Precision Emulation Approach to the GPU Acceleration of Ab Initio Electronic Structure Calculations

This study explores the use of INT8-based emulation for accelerating traditional FP64-based HPC workloads on modern GPU architectures. Through SCILIB-Accel automatic BLAS offload tool for cache-coherent Unified Memory Architecture, we emulate FP64 matrix multiplications in the LSMS CPU application in the MuST suite without code changes. We find that accuracy depends on both arithmetic precision and the properties of the operator, which can be dealt with through tunable precision emulation. Unlike traditional mixed-precision approaches, this method preserves original algorithms while optimizing hardware utilization. We showcase the potential of improving accuracy and performance at the same time. This work highlights the potential of AI-driven hardware to transform HPC, advocating for adaptive precision strategies in future scientific computing.

cs.DC↗

arXiv2026

From Misclassifications to Outliers: Joint Reliability Assessment in Classification

Building reliable classifiers is a fundamental challenge for deploying machine learning in real-world applications. A reliable system should not only detect out-of-distribution (OOD) inputs but also anticipate in-distribution (ID) errors by assigning low confidence to potentially misclassified samples. Yet, most prior work treats OOD detection and failure prediction as separated problems, overlooking their closed connection. We argue that reliability requires evaluating them jointly. To this end, we propose a unified evaluation framework that integrates OOD detection and failure prediction, quantified by our new metrics DS-F1 and DS-AURC, where DS denotes double scoring functions. Experiments on the OpenOOD benchmark show that double scoring functions yield classifiers that are substantially more reliable than traditional single scoring approaches. Our analysis further reveals that OOD-based approaches provide notable gains under simple or far-OOD shifts, but only marginal benefits under more challenging near-OOD conditions. Beyond evaluation, we extend the reliable classifier SURE and introduce SURE+, a new approach that significantly improves reliability across diverse scenarios. Together, our framework, metrics, and method establish a new benchmark for trustworthy classification and offer practical guidance for deploying robust models in real-world settings. The source code is publicly available at https://github.com/Intellindust-AI-Lab/SUREPlus.

cs.CV↗

arXiv2025

Parallel Seismic Data Processing Performance with Cloud-based Storage

This article introduces a general processing framework to effectively utilize waveform data stored on modern cloud platforms. The focus is hybrid processing schemes where a local system drives processing. We show that downloading files and doing all processing locally is problematic even when the local system is a high-performance compute cluster. Benchmark tests with parallel processing show that approach always creates a bottleneck as the volume of data being handled increases with more processes pulling data. We find a hybrid model where processing to reduce the volume of data transferred from the cloud servers to the local system can dramatically improve processing time. Tests implemented with Massively Parallel Analysis System for Seismology (MsPASS) utilizing Amazon Web Service's Lamba service yield throughput comparable to processing day files on a local HPC file system. Given the ongoing migration of seismology data to cloud storage, our results show doing some or all processing on the cloud will be essential for any processing involving large volumes of data.

physics.geo-ph↗

arXiv2025

Density Prediction of Income Distribution Based on Mixed Frequency Data

Modeling large dependent datasets in modern time series analysis is a crucial research area. One effective approach to handle such datasets is to transform the observations into density functions and apply statistical methods for further analysis. Income distribution forecasting, a common application scenario, benefits from predicting density functions as it accounts for uncertainty around point estimates, leading to more informed policy formulation. However, predictive modeling becomes challenging when dealing with mixed-frequency data. To address this challenge, this paper introduces a mixed data sampling regression model for probability density functions (PDF-MIDAS). To mitigate variance inflation caused by high-frequency prediction variables, we utilize exponential Almon polynomials with fewer parameters to regularize the coefficient structure. Additionally, we propose an iterative estimation method based on quadratic programming and the BFGS algorithm. Simulation analyses demonstrate that as the sample size for estimating density functions and observation length increase, the estimator approaches the true value. Real data analysis reveals that compared to single-sequence prediction models, PDF-MIDAS incorporating high-frequency exogenous variables offers a wider range of application scenarios with superior fitting and prediction performance.

stat.ME↗

arXiv2025

A Review of Cloud Computing in Seismology

Seismology has entered the petabyte era, driven by decades of continuous recordings of broadband networks, the increase in nodal seismic experiments, and the recent emergence of Distributed Acoustic Sensing (DAS). This review explains how commercial clouds - AWS, Google Cloud, and Azure - by providing object storage, elastic compute, and managed databases, enable researchers to "bring the code to the data," thereby overcoming traditional HPC solutions' bandwidth and capacity limitations. After literature reviews of cloud concepts and their research applications in seismology, we illustrate the capacities of cloud-native workflows using two canonical end-to-end demonstrations: 1) ambient noise seismology and cross-correlation, and 2) earthquake detection, discrimination, and phase picking. Both workflows utilized S3 for streaming I/O and DocumentDB for provenance, demonstrating that cloud throughput can rival on-premises HPC at comparable costs, scanning 100 TBs to 1.3 PBs of seismic data in a few hours or days of processing. The review also discusses research and education initiatives, the reproducibility benefits of containers, and cost pitfalls (e.g., egress, I/O fees) of energy-intensive seismological research computing. While designing cloud pipelines remains non-trivial, partnerships with research software engineers enable converting domain code into scalable, automated, and environmentally conscious solutions for next-generation seismology.

physics.geo-ph↗

arXiv2025

A Global-scale Database of Seismic Phases from Cloud-based Picking at Petabyte Scale

We present the first global-scale database of 4.3 billion P- and S-wave picks extracted from 1.3 PB continuous seismic data via a cloud-native workflow. Using cloud computing services on Amazon Web Services, we launched ~145,000 containerized jobs on continuous records from 47,354 stations spanning 2002-2025, completing in under three days. Phase arrivals were identified with a deep learning model, PhaseNet, through an open-source Python ecosystem for deep learning, SeisBench. To visualize and gain a global understanding of these picks, we present preliminary results about pick time series revealing Omori-law aftershock decay, seasonal variations linked to noise levels, and dense regional coverage that will enhance earthquake catalogs and machine-learning datasets. We provide all picks in a publicly queryable database, providing a powerful resource for researchers studying seismicity around the world. This report provides insights into the database and the underlying workflow, demonstrating the feasibility of petabyte-scale seismic data mining on the cloud and of providing intelligent data products to the community in an automated manner.

physics.geo-ph↗

arXiv2025

Multiscale Roughness of Upper Mantle Discontinuities Inferred from the USArray: Dependence on Tomography Models

We used 3D plane wave migration to image topography of the upper mantle discontinuities at 410 km and 660 km depth that defines the transition zone. We find both discontinuities have topography variation at all resolvable scales. In this paper we examine the dependency of discontinuity roughness on tomography models. We migrated a common USArray data set with a selection of regional, global, surface wave, and body wave tomography models to distinguish different scales of roughness. The objective is to appraise what features are potential artifact from using an inaccurate velocity model. We find that the largest-scale features depend on the choice of tomography model, while smaller-scale features appear to be almost completely independent of current generation models. We suggest that this observation is additional evidence of the existence of small-scale roughness on the upper mantle discontinuities not captured with the data sampling density of the USArray. We find all models based only on body wave travel times alone do not remove a continent scale offset of both discontinuities that correlates with the edge of the craton. We conclude that offset is an artifact linked to underestimation of wave speed in the upper mantle by pure body wave models. Models produced by joint inversion with surface wave dispersion data are less prone to this apparent artifact. The large-scale topography variation is consistent with rigid plate motion models of the subducted Farallon slab underneath North American. Smaller scale topography is found to have larger variation in regions where the vertical mantle flow through the transition zone is implied by transition zone thickness.

physics.geo-ph↗

arXiv2025

Automatic BLAS Offloading on Unified Memory Architecture: A Study on NVIDIA Grace-Hopper

Porting codes to GPU often requires major efforts. While several tools exist for automatically offload numerical libraries such as BLAS and LAPACK, they often prove impractical due to the high cost of mandatory data transfer. The new unified memory architecture in NVIDIA Grace-Hopper allows high bandwidth cache-coherent memory access of all memory from both CPU and GPU, potentially eliminating bottleneck faced in conventional architecture. This breakthrough opens up new avenues for application development and porting strategies. In this study, we introduce a new tool for automatic BLAS offload, the tool leverages the high speed cache coherent NVLink C2C interconnect in Grace-Hopper, and enables performant GPU offload for BLAS heavy applications with no code changes or recompilation. The tool was tested on two quantum chemistry or physics codes, great performance benefits were observed.

cs.DC↗

arXiv2025

A Pilot Study on Tunable Precision Emulation via Automatic BLAS Offloading

This study explores the use of automatic BLAS offloading and INT8-based emulation for accelerating traditional HPC workloads on modern GPU architectures. Through the use of low-bitwidth integer units and cache-coherent Unified Memory Architecture, we emulate double-precision matrix multiplications in the MuST application without code changes. We find that accuracy depends on both arithmetic precision and the properties of the operator, which can be dealt with through tunable precision emulation. Unlike traditional mixed-precision approaches, this method preserves original algorithms while optimizing hardware utilization. We showcases the potential of improving accuracy and performance at the same time. This work highlights the potential of AI-driven hardware to transform HPC, advocating for adaptive precision strategies in future scientific computing.

cs.DC↗

arXiv2024

When is truncated stop loss optimal?

The paper examines how reinsurance can be used to strike a balance between expected profit and VaR/CVaR risk. Conditions making truncated stop loss contracts optimal are derived, and it is argued that those are usually satisfied in practice. One of the prerequisites is that reinsurance is not too cheap, and an argument resembling arbitrage suggests that it is not.

stat.AP↗

arXiv2023

Fine-grained Policy-driven I/O Sharing for Burst Buffers

A burst buffer is a common method to bridge the performance gap between the I/O needs of modern supercomputing applications and the performance of the shared file system on large-scale supercomputers. However, existing I/O sharing methods require resource isolation, offline profiling, or repeated execution that significantly limit the utilization and applicability of these systems. Here we present ThemisIO, a policy-driven I/O sharing framework for a remote-shared burst buffer: a dedicated group of I/O nodes, each with a local storage device. ThemisIO preserves high utilization by implementing opportunity fairness so that it can reallocate unused I/O resources to other applications. ThemisIO accurately and efficiently allocates I/O cycles among applications, purely based on real-time I/O behavior without requiring user-supplied information or offline-profiled application characteristics. ThemisIO supports a variety of fair sharing policies, such as user-fair, size-fair, as well as composite policies, e.g., group-then-user-fair. All these features are enabled by its statistical token design. ThemisIO can alter the execution order of incoming I/O requests based on assigned tokens to precisely balance I/O cycles between applications via time slicing, thereby enforcing processing isolation. Experiments using I/O benchmarks show that ThemisIO sustains 13.5-13.7% higher I/O throughput and 19.5-40.4% lower performance variation than existing algorithms. For real applications, ThemisIO significantly reduces the slowdown by 59.1-99.8% caused by I/O interference.

cs.DC↗

arXiv2019

Optimal reinsurance for risk over surplus ratios

Optimal reinsurance when Value at Risk and expected surplus is balanced through their ratio is studied, and it is demonstrated how results for risk-adjusted surplus can be utilized. Simplifications for large portfolios are derived, and this large-portfolio study suggests a new condition on the reinsurance pricing regime which is crucial for the results obtained. One or two-layer contracts now become optimal for both risk-adjusted surplus and the risk over expected surplus ratio, but there is no second layer when portfolios are large or when reinsurance prices are below some threshold. Simple approximations of the optimum portfolio are considered, and their degree of degradation compared to the optimum is studied which leads to theoretical degradation rates as the number of policies grows. The theory is supported by numerical experiments which suggest that the shape of the claim severity distributions may not be of primary importance when designing an optimal reinsurance program. It is argued that the approach can be applied to Conditional Value at Risk as well.

stat.AP↗

arXiv2019

How much is optimal reinsurance degraded by error?

The literature on optimal reinsurance does not deal with how much the effectiveness of such solutions is degraded by errors in parameters and models. The issue is investigated through both asymptotics and numerical studies. It is shown that the rate of degradation is often $O(1/n)$ as the sample size $n$ of historical observations becomes infinite. Criteria based on Value at Risk are exceptions that may achieve only $O(1/\sqrt{n})$. These theoretical results are supported by numerical studies. A Bayesian perspective on how to integrate risk caused by parameter error is offered as well.

stat.AP↗

arXiv2019

Modelling extreme claims via composite models and threshold selection methods

The existence of large and extreme claims of a non-life insurance portfolio influences the ability of (re)insurers to estimate the reserve. The excess over-threshold method provides a way to capture and model the typical behaviour of insurance claim data. This paper discusses several composite models with commonly used bulk distributions, combined with a 2-parameter Pareto distribution above the threshold. We have explored how several threshold selection methods perform when estimating the reserve as well as the effect of the choice of bulk distribution, with varying sample size and tail properties. To investigate this, a simulation study has been performed. Our study shows that when data are sufficient, the square root rule has the overall best performance in terms of the quality of the reserve estimate. The second best is the exponentiality test, especially when the right tail of the data is extreme. As the sample size becomes small, the simultaneous estimation has the best performance. Further, the influence of the choice of bulk distribution seems to be rather large, especially when the distribution is heavy-tailed. Moreover, it shows that the empirical estimate of $p_{\leq b}$, the probability that a claim is below the threshold, is more robust than the theoretical one.

stat.AP↗