arXiv subjects

Sidharth Kumar

Publications and source records attributed to Sidharth Kumar.

At least 19 recordsLinked to original sources

arXiv2026

Configurable and Hierarchical Allreduce

MPI_Allreduce is among the most performance-critical collectives in large-scale scientific computing and distributed machine learning, yet the small- and medium-message regime remains challenging: latency, synchronization depth, and strong hardware hierarchy between intra- and inter-domain communication all compound per-invocation cost. We present CHIARA, a configurable hierarchical Allreduce that encodes hardware hierarchy through a logical batch-lane topology and executes a staged schedule in which only a bounded portion of the reduction vector is active at a time. Inter-batch communication is distributed across multiple ranks via a rotating-root lane primitive, avoiding centralized leaders. Tool further enables a semi-composed Rabenseifner-style Allreduce by preserving a lane-aligned intermediate layout across the Reduce-Scatter/Allgather boundary, eliminating redundant intra-domain reorganization. We evaluate Tool on Polaris, Aurora, and Fugaku, achieving speedups of up to 1.94x, 13.43x, and 13.48x over vendor MPI_Allreduce, and up to 2.2x end-to-end speedup in a parallel k-means application.

cs.DC↗

arXiv2026

Terascale Query Processing in the Browser: Rethinking GPU Acceleration

Recursive query computation, central to graph algorithms and relational databases, demands GPU acceleration due to its inherent computational intensity. While substantial prior work addresses GPU implementations of recursive queries that require fixed-point evaluation, existing systems are restricted to native execution environments. We introduce WGLog, the first web-browser-native GPU engine for compute-bound recursive database queries. WGLog is built entirely on WebGPU compute shaders, a cross-platform API that enables GPU acceleration in web browsers. WGLog leverages two key technical innovations. First, we replace hash-table-based joins with atomic-free sorted-array joins, eliminating the serialization bottleneck that hash tables suffer on skewed graphs. Second, we develop an asynchronous execution pipeline using WebGPU's indirect dispatch capability, which eliminates GPU-host synchronizations that would otherwise dominate per-iteration overhead. On representative workloads, WGLog delivers a 1.48--4.68x speedup over native GPU systems and orders-of-magnitude improvement over CPU and WebAssembly implementations.

cs.DB↗

arXiv2026

Efficient Compression of Structured and Unstructured Volumes via Learned 3D Gaussian Representation

Recent work has shown that implicit neural representations (INRs) can be trained to effectively compress structured and unstructured volume data, allowing for direct data querying with a reduced memory footprint. However, as existing INRs for unstructured volumes do not encode geometry, they require partial mesh storage for later sampling, limiting achievable compression. At the same time, novel view synthesis methods have shown that explicit collections of 3D Gaussians can be used to accurately visualize volume data. In this work, we introduce an explicit model for volume data compression based on 3D Gaussian primitives. We reinterpret collections of 3D Gaussians as an explicit representation of a scalar field and use a sampling strategy that reconstructs scalar values at spatial locations through weighted aggregation of intersecting Gaussians. We develop optimized CUDA-accelerated pipelines for structured and unstructured model sampling, loss functions that encourage accurate domain encoding by our models, and a novel sampling-error based densification strategy. Our explicit formulation naturally encodes domain geometry, eliminating the need for mesh storage in unstructured volumes and introducing significantly higher compression opportunities. Compared to existing INRs, we demonstrate that our explicit model achieves competitive reconstruction quality with significant training speedups on structured volumes, while markedly outperforming in all metrics on unstructured volumes.

cs.LG↗

arXiv2026

Scaling Worst-Case Optimal Datalog to GPUs

Datalog is a declarative logic-programming language used for complex analytic reasoning workloads such as program analysis and graph analytics. Datalog's popularity is due to its unique price-point, marrying logic-defined specification with the potential for massive data parallelism. While traditional engines are CPU-based, the memory-bound nature of Datalog has led to increasing interest in leveraging GPUs. These engines beat CPU-based engines by operationalizing iterated relational joins via SIMT-friendly join algorithms. Unfortunately, all existing GPU Datalog engines are built on binary joins, which are inadequate for the complex multi-way queries arising in production systems such as DOOP and ddisasm. For these queries, binary decomposition can incur the AGM bound asymptotic blowup in time and space, leading to OOM failures regardless of join order. Worst-Case Optimal Joins (WCOJ) avoid this blowup, but their attribute-at-a-time intersections map poorly to SIMT hardware under key skew, causing severe load imbalance across Streaming Multiprocessors (SMs). We present SRDatalog, the first GPU Datalog engine based on WCOJ. SRDatalog uses flat columnar storage and two-phase deterministic memory allocation to avoid the OOM failures of binary joins and the index-rebuild overheads of static WCOJ systems. To mitigate skew and hide hardware stalls, SRDatalog further employs root-level histogram-guided load balancing, structural helper-relation splitting, and stream-aligned rule multiplexing. On real-world program-analysis workloads, SRDatalog achieves geometric-mean speedups of 21x to 47x.

cs.DB↗

arXiv2026

Accelerating Stroke MRI with Diffusion Probabilistic Models through Large-Scale Pre-training and Target-Specific Fine-Tuning

Purpose: To develop a data-efficient strategy for accelerated MRI reconstruction with Diffusion Probabilistic Generative Models (DPMs) that enables faster scan times in clinical stroke MRI when only limited fully-sampled data samples are available. Methods: Our simple training strategy, inspired by the foundation model paradigm, first trains a DPM on a large, diverse collection of publicly available brain MRI data in fastMRI and then fine-tunes on a small dataset from the target application using carefully selected learning rates and fine-tuning durations. The approach is evaluated on controlled fastMRI experiments and on clinical stroke MRI data with a blinded clinical reader study. Results: DPMs pre-trained on approximately 4000 subjects with non-FLAIR contrasts and fine-tuned on FLAIR data from only 20 target subjects achieve reconstruction performance comparable to models trained with substantially more target-domain FLAIR data across multiple acceleration factors. Experiments reveal that moderate fine-tuning with a reduced learning rate yields improved performance, while insufficient or excessive fine-tuning degrades reconstruction quality. When applied to clinical stroke MRI, a blinded reader study involving two neuroradiologists indicates that images reconstructed using the proposed approach from $2 \times$ accelerated data are non-inferior to standard-of-care in terms of image quality and structural delineation. Conclusion: Large-scale pre-training combined with targeted fine-tuning enables DPM-based MRI reconstruction in data-constrained, accelerated clinical stroke MRI. The proposed approach substantially reduces the need for large application-specific datasets while maintaining clinically acceptable image quality, supporting the use of foundation-inspired diffusion models for accelerated MRI in targeted applications.

eess.IV↗

arXiv2025

Volume Encoding Gaussians: Transfer Function-Agnostic 3D Gaussians for Volume Rendering

Visualizing the large-scale datasets output by HPC resources presents a difficult challenge, as the memory and compute power required become prohibitively expensive for end user systems. Novel view synthesis techniques can address this by producing a small, interactive model of the data, requiring only a set of training images to learn from. While these models allow accessible visualization of large data and complex scenes, they do not provide the interactions needed for scientific volumes, as they do not support interactive selection of transfer functions and lighting parameters. To address this, we introduce Volume Encoding Gaussians (VEG), a 3D Gaussian-based representation for volume visualization that supports arbitrary color and opacity mappings. Unlike prior 3D Gaussian Splatting (3DGS) methods that store color and opacity for each Gaussian, VEG decouple the visual appearance from the data representation by encoding only scalar values, enabling transfer function-agnostic rendering of 3DGS models. To ensure complete scalar field coverage, we introduce an opacity-guided training strategy, using differentiable rendering with multiple transfer functions to optimize our data representation. This allows VEG to preserve fine features across the full scalar range of a dataset while remaining independent of any specific transfer function. Across a diverse set of volume datasets, we demonstrate that our method outperforms the state-of-the-art on transfer functions unseen during training, while requiring a fraction of the memory and training time.

cs.GR↗

arXiv2025

Column-Oriented Datalog on the GPU

Datalog is a logic programming language widely used in knowledge representation and reasoning (KRR), program analysis, and social media mining due to its expressiveness and high performance. Traditionally, Datalog engines use either row-oriented or column-oriented storage. Engines like VLog and Nemo favor column-oriented storage for efficiency on limited-resource machines, while row-oriented engines like Souffle use advanced data structures with locking to perform better on multi-core CPUs. The advent of modern datacenter GPUs, such as the NVIDIA H100 with its ability to run over 16k threads simultaneously and high memory bandwidth, has reopened the debate on which storage layout is more effective. This paper presents the first column-oriented Datalog engines tailored to the strengths of modern GPUs. We present VFLog, a CUDA-based Datalog runtime library with a column-oriented GPU datastructure that supports all necessary relational algebra operations. Our results demonstrate over 200x performance gains over SOTA CPU-based column-oriented Datalog engines and a 2.5x speedup over GPU Datalog engines in various workloads, including KRR.

cs.DB↗

arXiv2024

Datalog with First-Class Facts

Datalog is a popular logic programming language for deductive reasoning tasks in a wide array of applications, including business analytics, program analysis, and ontological reasoning. However, Datalog's restriction to flat facts over atomic constants leads to challenges in working with tree-structured data, such as derivation trees or abstract syntax trees. To ameliorate Datalog's restrictions, popular extensions of Datalog support features such as existential quantification in rule heads (Datalog$^\pm$, Datalog$^\exists$) or algebraic data types (Souffl\'e). Unfortunately, these are imperfect solutions for reasoning over structured and recursive data types, with general existentials leading to complex implementations requiring unification, and ADTs unable to trigger rule evaluation and failing to support efficient indexing. We present DL$^{\exists!}$, a Datalog with first-class facts, wherein every fact is identified with a Skolem term unique to the fact. We show that this restriction offers an attractive price point for Datalog-based reasoning over tree-shaped data, demonstrating its application to databases, artificial intelligence, and programming languages. We implemented DL$^{\exists!}$ as a system \slog{}, which leverages the uniqueness restriction of DL$^{\exists!}$ to enable a communication-avoiding, massively-parallel implementation built on MPI. We show that Slog outperforms leading systems (Nemo, Vlog, RDFox, and Souffl\'e) on a variety of benchmarks, with the potential to scale to thousands of threads.

cs.DB↗

arXiv2024

Robust multi-coil MRI reconstruction via self-supervised denoising

We study the effect of incorporating self-supervised denoising as a pre-processing step for training deep learning (DL) based reconstruction methods on data corrupted by Gaussian noise. K-space data employed for training are typically multi-coil and inherently noisy. Although DL-based reconstruction methods trained on fully sampled data can enable high reconstruction quality, obtaining large, noise-free datasets is impractical. We leverage Generalized Stein's Unbiased Risk Estimate (GSURE) for denoising. We evaluate two DL-based reconstruction methods: Diffusion Probabilistic Models (DPMs) and Model-Based Deep Learning (MoDL). We evaluate the impact of denoising on the performance of these DL-based methods in solving accelerated multi-coil magnetic resonance imaging (MRI) reconstruction. The experiments were carried out on T2-weighted brain and fat-suppressed proton-density knee scans. We observed that self-supervised denoising enhances the quality and efficiency of MRI reconstructions across various scenarios. Specifically, employing denoised images rather than noisy counterparts when training DL networks results in lower normalized root mean squared error (NRMSE), higher structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR) across different SNR levels, including 32dB, 22dB, and 12dB for T2-weighted brain data, and 24dB, 14dB, and 4dB for fat-suppressed knee data. Overall, we showed that denoising is an essential pre-processing technique capable of improving the efficacy of DL-based MRI reconstruction methods under diverse conditions. By refining the quality of input data, denoising enables training more effective DL networks, potentially bypassing the need for noise-free reference MRI scans.

eess.IV↗

arXiv2024

Configurable Non-uniform All-to-all Algorithms

MPI_Alltoallv generalizes the uniform all-to-all communication (MPI_Alltoall) by enabling the exchange of data blocks of varied sizes among processes. This function plays a crucial role in many applications, such as FFT computation and relational algebra operations. Popular MPI libraries, such as MPICH and OpenMPI, implement MPI_Alltoall using a combination of linear and logarithmic algorithms. However, MPI_Alltoallv typically relies only on variations of linear algorithms, missing the benefits of logarithmic approaches. Furthermore, current algorithms also overlook the intricacies of modern HPC system architectures, such as the significant performance gap between intra-node (local) and inter-node (global) communication. This paper introduces a set of Tunable Non-uniform All-to-all algorithms, denoted TuNA{l}{g}, where g and l refer to global (inter-node) and local (intra-node) communication hierarchies.These algorithms consider key factors such as the hierarchical architecture of HPC systems, network congestion, the number of data exchange rounds, and the communication burst size. The algorithm efficiently addresses the trade-off between bandwidth maximization and latency minimization that existing implementations struggle to optimize. We show a performance improvement over the state-of-the-art implementations by factors of 42x and 138x on Polaris and Fugaku, respectively.

cs.DC↗

arXiv2024

Enabling Fast and Accurate Crowdsourced Annotation for Elevation-Aware Flood Extent Mapping

Mapping the extent of flood events is a necessary and important aspect of disaster management. In recent years, deep learning methods have evolved as an effective tool to quickly label high resolution imagery and provide necessary flood extent mappings. These methods, though, require large amounts of annotated training data to create models that are accurate and robust to new flooded imagery. In this work, we present FloodTrace, a web-based application that enables effective crowdsourcing of flooded region annotation for machine learning applications. To create this application, we conducted extensive interviews with domain experts to produce a set of formal requirements. Our work brings topological segmentation tools to the web and greatly improves annotation efficiency compared to the state-of-the-art. The user-friendliness of our solution allows researchers to outsource annotations to non-experts and utilize them to produce training data with equal quality to fully expert-labeled data. We conducted a user study to confirm the effectiveness of our application in which 266 graduate students annotated high-resolution aerial imagery from Hurricane Matthew in North Carolina. Experimental results show the efficiency benefits of our application for untrained users, with median annotation time less than half the state-of-the-art annotation method. In addition, using our aggregation and correction framework, flood detection models trained on crowdsourced annotations were able to achieve performance equal to models trained on fully expert-labeled annotations, while requiring a fraction of the time on the part of the expert.

cs.CV↗

arXiv2024

Ambient Diffusion Posterior Sampling: Solving Inverse Problems with Diffusion Models Trained on Corrupted Data

We provide a framework for solving inverse problems with diffusion models learned from linearly corrupted data. Firstly, we extend the Ambient Diffusion framework to enable training directly from measurements corrupted in the Fourier domain. Subsequently, we train diffusion models for MRI with access only to Fourier subsampled multi-coil measurements at acceleration factors R= 2,4,6,8. Secondly, we propose Ambient Diffusion Posterior Sampling (A-DPS), a reconstruction algorithm that leverages generative models pre-trained on one type of corruption (e.g. image inpainting) to perform posterior sampling on measurements from a different forward process (e.g. image blurring). For MRI reconstruction in high acceleration regimes, we observe that A-DPS models trained on subsampled data are better suited to solving inverse problems than models trained on fully sampled data. We also test the efficacy of A-DPS on natural image datasets (CelebA, FFHQ, and AFHQ) and show that A-DPS can sometimes outperform models trained on clean data for several image restoration tasks in both speed and performance.

cs.CV↗

arXiv2023

Optimizing Datalog for the GPU

Modern Datalog engines (e.g., LogicBlox, Souffl\'e, ddlog) enable their users to write declarative queries which compute recursive deductions over extensional facts, leaving high-performance operationalization (query planning, semi-na\"ive evaluation, and parallelization) to the engine. Such engines form the backbone of modern high-throughput applications in static analysis, network monitoring, and social-media mining. In this paper, we present a methodology for implementing a modern in-memory Datalog engine on data center GPUs, allowing us to achieve significant (up to 45x) gains compared to Souffl\'e (a modern CPU-based engine) on context-sensitive points-to analysis of httpd. We present GPUlog, a Datalog engine backend that implements iterated relational algebra kernels over a novel range-indexed data structure we call the hash-indexed sorted array (HISA). HISA combines the algorithmic benefits of incremental range-indexed relations with the raw computation throughput of operations over dense data structures. Our experiments show that GPUlog is significantly faster than CPU-based Datalog engines while achieving a favorable memory footprint compared to contemporary GPU-based joins.

cs.DB↗

arXiv2023

Speculative Progressive Raycasting for Memory Constrained Isosurface Visualization of Massive Volumes

New web technologies have enabled the deployment of powerful GPU-based computational pipelines that run entirely in the web browser, opening a new frontier for accessible scientific visualization applications. However, these new capabilities do not address the memory constraints of lightweight end-user devices encountered when attempting to visualize the massive data sets produced by today's simulations and data acquisition systems. In this paper, we propose a novel implicit isosurface rendering algorithm for interactive visualization of massive volumes within a small memory footprint. We achieve this by progressively traversing a wavefront of rays through the volume and decompressing blocks of the data on-demand to perform implicit ray-isosurface intersections. The progressively rendered surface is displayed after each pass to improve interactivity. Furthermore, to accelerate rendering and increase GPU utilization, we introduce speculative ray-block intersection into our algorithm, where additional blocks are traversed and intersected speculatively along rays as other rays terminate to exploit additional parallelism in the workload. Our entire pipeline is run in parallel on the GPU to leverage the parallel computing power that is available even on lightweight end-user devices. We compare our algorithm to the state of the art in low-overhead isosurface extraction and demonstrate that it achieves 1.7x-5.7x reductions in memory overhead and up to 8.4x reductions in data decompressed.

cs.GR↗

arXiv2023

Solving Inverse Problems with Score-Based Generative Priors learned from Noisy Data

We present SURE-Score: an approach for learning score-based generative models using training samples corrupted by additive Gaussian noise. When a large training set of clean samples is available, solving inverse problems via score-based (diffusion) generative models trained on the underlying fully-sampled data distribution has recently been shown to outperform end-to-end supervised deep learning. In practice, such a large collection of training data may be prohibitively expensive to acquire in the first place. In this work, we present an approach for approximately learning a score-based generative model of the clean distribution, from noisy training data. We formulate and justify a novel loss function that leverages Stein's unbiased risk estimate to jointly denoise the data and learn the score function via denoising score matching, while using only the noisy samples. We demonstrate the generality of SURE-Score by learning priors and applying posterior sampling to ill-posed inverse problems in two practical applications from different domains: compressive wireless multiple-input multiple-output channel estimation and accelerated 2D multi-coil magnetic resonance imaging reconstruction, where we demonstrate competitive reconstruction performance when learning at signal-to-noise ratio values of 0 and 10 dB, respectively.

cs.LG↗

arXiv2022

Higher-Order, Data-Parallel Structured Deduction

State-of-the-art Datalog engines include expressive features such as ADTs (structured heap values), stratified aggregation and negation, various primitive operations, and the opportunity for further extension using FFIs. Current parallelization approaches for state-of-art Datalogs target shared-memory locking data-structures using conventional multi-threading, or use the map-reduce model for distributed computing. Furthermore, current state-of-art approaches cannot scale to formal systems which pervasively manipulate structured data due to their lack of indexing for structured data stored in the heap. In this paper, we describe a new approach to data-parallel structured deduction that involves a key semantic extension of Datalog to permit first-class facts and higher-order relations via defunctionalization, an implementation approach that enables parallelism uniformly both across sets of disjoint facts and over individual facts with nested structure. We detail a core language, $DL_s$, whose key invariant (subfact closure) ensures that each subfact is materialized as a top-class fact. We extend $DL_s$ to Slog, a fully-featured language whose forms facilitate leveraging subfact closure to rapidly implement expressive, high-performance formal systems. We demonstrate Slog by building a family of control-flow analyses from abstract machines, systematically, along with several implementations of classical type systems (such as STLC and LF). We performed experiments on EC2, Azure, and ALCF's Theta at up to 1000 threads, showing orders-of-magnitude scalability improvements versus competing state-of-art systems.

cs.PL↗

arXiv2022

Accelerated Motion Correction with Deep Generative Diffusion Models

Magnetic Resonance Imaging (MRI) is a powerful medical imaging modality, but unfortunately suffers from long scan times which, aside from increasing operational costs, can lead to image artifacts due to patient motion. Motion during the acquisition leads to inconsistencies in measured data that manifest as blurring and ghosting if unaccounted for in the image reconstruction process. Various deep learning based reconstruction techniques have been proposed which decrease scan time by reducing the number of measurements needed for a high fidelity reconstructed image. Additionally, deep learning has been used to correct motion using end-to-end techniques. This, however, increases susceptibility to distribution shifts at test time (sampling pattern, motion level). In this work we propose a framework for jointly reconstructing highly sub-sampled MRI data while estimating patient motion using diffusion based generative models. Our method does not make specific assumptions on the sampling trajectory or motion pattern at training time and thus can be flexibly applied to various types of measurement models and patient motion. We demonstrate our framework on retrospectively accelerated 2D brain MRI corrupted by rigid motion.

eess.IV↗

arXiv2022

FSE Compensated Motion Correction for MRI Using Data Driven Methods

Magnetic Resonance Imaging (MRI) is a widely used medical imaging modality boasting great soft tissue contrast without ionizing radiation, but unfortunately suffers from long acquisition times. Long scan times can lead to motion artifacts, for example due to bulk patient motion such as head movement and periodic motion produced by the heart or lungs. Motion artifacts can degrade image quality and in some cases render the scans nondiagnostic. To combat this problem, prospective and retrospective motion correction techniques have been introduced. More recently, data driven methods using deep neural networks have been proposed. As a large number of publicly available MRI datasets are based on Fast Spin Echo (FSE) sequences, methods that use them for training should incorporate the correct FSE acquisition dynamics. Unfortunately, when simulating training data, many approaches fail to generate accurate motion-corrupt images by neglecting the effects of the temporal ordering of the k-space lines as well as neglecting the signal decay throughout the FSE echo train. In this work, we highlight this consequence and demonstrate a training method which correctly simulates the data acquisition process of FSE sequences with higher fidelity by including sample ordering and signal decay dynamics. Through numerical experiments, we show that accounting for the FSE acquisition leads to better motion correction performance during inference.

eess.SP↗