Arun K. Somani
YOU?
Author Swipe
View article: Where Does mBERT Understand Code-Mixing? Layer-Dependent Performance on Semantic Tasks
Where Does mBERT Understand Code-Mixing? Layer-Dependent Performance on Semantic Tasks Open
Semantic tasks like lexical relation prediction and word analogy are crucial for deep language understanding, yet pose significant challenges when applied to code-mixed text, where multiple languages are interwoven. This study investigates…
View article: CQS-Attention: Scaling Up the Standard Attention Computation for Infinitely Long Sequences
CQS-Attention: Scaling Up the Standard Attention Computation for Infinitely Long Sequences Open
Transformer models suffer from unaffordable high memory consumption when the sequence is long and standard self-attention is utilized. We developed a sequence parallelism scheme called CQS-Attention that can break the limit of sequence len…
View article: SPLAS: An Autonomous Lightweight and Versatile Image Classification Robustness Benchmark
SPLAS: An Autonomous Lightweight and Versatile Image Classification Robustness Benchmark Open
Image classification is one of the most fundamental vision tasks and remarkable progress in prediction accuracy has been made with deep neural networks and vision transformers in the last decade. However, in real-world applications, these …
View article: ConVision Benchmark: A Contemporary Framework to Benchmark CNN and ViT Models
ConVision Benchmark: A Contemporary Framework to Benchmark CNN and ViT Models Open
Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) have shown remarkable performance in computer vision tasks, including object detection and image recognition. These models have evolved significantly in architecture, effi…
View article: A survey of techniques for optimizing transformer inference
A survey of techniques for optimizing transformer inference Open
View article: A Survey of Techniques for Optimizing Transformer Inference
A Survey of Techniques for Optimizing Transformer Inference Open
Recent years have seen a phenomenal rise in performance and applications of transformer neural networks. The family of transformer networks, including Bidirectional Encoder Representations from Transformer (BERT), Generative Pretrained Tra…
View article: Improvement and Evaluation of Resilience of Adaptive Cruise Control Against Spoofing Attacks Using Intrusion Detection System
Improvement and Evaluation of Resilience of Adaptive Cruise Control Against Spoofing Attacks Using Intrusion Detection System Open
The Adaptive Cruise Control (ACC) system automatically adjusts the vehicle speed to maintain a safe distance between the vehicle and the lead (ahead) vehicle. The controller's decision to accelerate or decelerate is computed using the targ…
View article: Dynamic advance reservation with delayed allocation
Dynamic advance reservation with delayed allocation Open
A method of scheduling data transmissions from a source to a destination, includes the steps of: providing a communication system having a number of channels and a number of paths, each of the channels having a plurality of designated time…
View article: Neural Architecture Search Benchmarks: Insights and Survey
Neural Architecture Search Benchmarks: Insights and Survey Open
Neural Architecture Search (NAS), a promising and fast-moving research field, aims to automate the architectural design of Deep Neural Networks (DNNs) to achieve better performance on the given task and dataset. NAS methods have been very …
View article: Differentiable Neural Architecture, Mixed Precision and Accelerator Co-Search
Differentiable Neural Architecture, Mixed Precision and Accelerator Co-Search Open
Quantization, effective Neural Network architecture, and efficient accelerator hardware are three important design paradigms to maximize accuracy and efficiency. Mixed Precision Quantization is a process of assigning different precision to…
View article: Efficient Design Space Exploration for Sparse Mixed Precision Neural Architectures
Efficient Design Space Exploration for Sparse Mixed Precision Neural Architectures Open
Pruning and Quantization are two effective Deep Neural Network (DNN) compression methods for efficient inference on various hardware platforms. Pruning refers to removing unimportant weights or nodes, whereas Quantization converts the floa…
View article: William C. Carter Award PhD Dissertation Award in Dependability
William C. Carter Award PhD Dissertation Award in Dependability Open
The
View article: Neural Architecture Search Survey: A Hardware Perspective
Neural Architecture Search Survey: A Hardware Perspective Open
We review the problem of automating hardware-aware architectural design process of Deep Neural Networks (DNNs). The field of Convolutional Neural Network (CNN) algorithm design has led to advancements in many fields, such as computer visio…
View article: An Effective Two-stage Noise Training Methodology for Classification of Breast Ultrasound Images
An Effective Two-stage Noise Training Methodology for Classification of Breast Ultrasound Images Open
View article: Neural Architecture Search for Transformers: A Survey
Neural Architecture Search for Transformers: A Survey Open
Transformer-based Deep Neural Network architectures have gained tremendous interest due to their effectiveness in various applications across Natural Language Processing (NLP) and Computer Vision (CV) domains. These models are the de facto…
View article: ARA
ARA Open
The rural US includes 72% of the nation's land and 46 million people, and it serves as major sources of food and energy for the nation. Thus rural prosperity is essential to US wellbeing. As a foundation for next-generation rural economy a…
View article: Physical Wireless Resource Virtualization for Software-Defined Whole-Stack Slicing
Physical Wireless Resource Virtualization for Software-Defined Whole-Stack Slicing Open
Radio access network (RAN) virtualization is gaining more and more ground and expected to re-architect the next-generation cellular networks. Existing RAN virtualization studies and solutions have mostly focused on sharing communication ca…
View article: Addressing multiple bit/symbol errors in DRAM subsystem
Addressing multiple bit/symbol errors in DRAM subsystem Open
As DRAM technology continues to evolve towards smaller feature sizes and increased densities, faults in DRAM subsystem are becoming more severe. Current servers mostly use CHIPKILL based schemes to tolerate up-to one/two symbol errors per …
View article: Physical Wireless Resource Virtualization for Software-Defined\n Whole-Stack Slicing
Physical Wireless Resource Virtualization for Software-Defined\n Whole-Stack Slicing Open
Radio access network (RAN) virtualization is gaining more and more ground and\nexpected to re-architect the next-generation cellular networks. Existing RAN\nvirtualization studies and solutions have mostly focused on sharing\ncommunication…
View article: High Speed Systems Engineering: A New Trend In Electrical And Computer Engineering
High Speed Systems Engineering: A New Trend In Electrical And Computer Engineering Open
This paper introduces the main ideas and program objectives for High Speed Systems Engineering (HSSE). HSSE (funded by NSF CISE-EAI)1 has been proposed by our team as a new and viable platform for integrating engineering education, researc…
View article: AUTOMATED SHAPE-BASED PAVEMENT CRACK DETECTION APPROACH
AUTOMATED SHAPE-BASED PAVEMENT CRACK DETECTION APPROACH Open
Pavements are critical man-made infrastructure systems that undergo repeated traffic and environmental loadings. Consequently, they deteriorate with time and manifest certain distresses. To ensure long-lasting performance and appropriate l…
View article: Fast Document Similarity Computations using GPGPU
Fast Document Similarity Computations using GPGPU Open
View article: Enhanced Feature Mining and Classifier Models to Predict Customer Churn for an e-Retailer
Enhanced Feature Mining and Classifier Models to Predict Customer Churn for an e-Retailer Open
e-Commerce (electronic commerce, or EC) include the buying, selling of goods and services and the transmitting of funds or data, over an electronic network. These business transactions occur business-to-business (B2B), business-to-consumer…
View article: Big Data Cluster Analysis: A Study of Existing Techniques and Future Directions
Big Data Cluster Analysis: A Study of Existing Techniques and Future Directions Open
Cluster analysis (clustering) is a fundamental problem in an unsupervised machine learning domain. It has a huge range of applications in various fields, including bioinformatics, gene sequencing, market basket research, medicine, social n…
View article: Understanding the Data Science behind Business Analytics
Understanding the Data Science behind Business Analytics Open
Big Data analytics has become the engine for business analytics today. Companies are using Big Data to analyze their business processes, formulate future business strategies and, extensively, employ it for decision making. Companies such a…
View article: On-Disk Data Processing: Issues and Future Directions
On-Disk Data Processing: Issues and Future Directions Open
In this paper, we present a survey of "on-disk" data processing (ODDP). ODDP, which is a form of near-data processing, refers to the computing arrangement where the secondary storage drives have the data processing capability. Proposed ODD…
View article: Host managed contention avoidance storage solutions for Big Data
Host managed contention avoidance storage solutions for Big Data Open
The performance gap between compute and storage is fairly considerable. This results in a mismatch between the application needs from storage and what storage can deliver. The full potential of storage devices cannot be harnessed till all …
View article: Managing contamination delay to improve Timing Speculation architectures
Managing contamination delay to improve Timing Speculation architectures Open
Timing Speculation (TS) is a widely known method for realizing better-than-worst-case systems. Aggressive clocking, realizable by TS, enable systems to operate beyond specified safe frequency limits to effectively exploit the data dependen…
View article: Scaling Distributed All-Pairs Algorithms: Manage Computation and Limit Data Replication with Quorums
Scaling Distributed All-Pairs Algorithms: Manage Computation and Limit Data Replication with Quorums Open
In this paper we propose and prove that cyclic quorum sets can efficiently manage all-pairs computations and data replication. The quorums are O(N/sqrt(P)) in size, up to 50% smaller than the dual N/sqrt(P) array implementations, and signi…
View article: Unidirectional Quorum-Based Cycle Planning for Efficient Resource Utilization and Fault-Tolerance
Unidirectional Quorum-Based Cycle Planning for Efficient Resource Utilization and Fault-Tolerance Open
In this paper, we propose a greedy cycle direction heuristic to improve the generalized $\mathbf{R}$ redundancy quorum cycle technique. When applied using only single cycles rather than the standard paired cycles, the generalized $\mathbf{…