- Topic Modeling
- Expert finding and Q&A systems
- Advanced Data Storage Technologies
- Advanced Wireless Communication Techniques
- Interconnection Networks and Systems
- Wireless Communication Networks Research
- Parallel Computing and Optimization Techniques
- Advanced MIMO Systems Optimization
- Graph Theory and Algorithms
- Digital Filter Design and Implementation
- Algorithms and Data Compression
- Wikis in Education and Collaboration
- Cognitive Radio Networks and Spectrum Sensing
- Advanced Wireless Network Optimization
- PAPR reduction in OFDM
- Handwritten Text Recognition Techniques
- Radio Frequency Integrated Circuit Design
- Error Correcting Code Techniques
- Electronic Health Records Systems
- Language, Metaphor, and Cognition
- Information Retrieval and Search Behavior
- Wireless Body Area Networks
- Electromagnetic Scattering and Analysis
- Blind Source Separation Techniques
- Advanced Adaptive Filtering Techniques
The University of Texas Health Science Center at Houston
2022-2023
University of Illinois Urbana-Champaign
2019-2021
Nokia (Finland)
2014-2017
Aalto University
2017
Tampere University
2010-2014
Tampere University of Applied Sciences
2010-2013
Software Defined Radio (SDR) is an innovative approach which becoming a more and promising technology for future mobile handsets. Several proposals in the field of embedded systems have been introduced by different universities industries to support SDR applications. This article presents overview current platforms analyzes related architectural choices, issues SDR, as well potential trends.
In this paper, we present an update to our previous submission on k-truss decomposition from Graph Challenge 2018. For single k implementation, propose multiple algorithmic optimizations that significantly improve performance by up 35.2x (6.9x average) compared GPU implementation. addition, a scalable multi-GPU implementation in which each handles different `k' value. Compared prior the proposed approach is faster 151.3x (78.8x average). case when edges with only maximal are sought,...
Omer Anjum, Hongyu Gong, Suma Bhat, Wen-Mei Hwu, JinJun Xiong. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.
An energy-efficient fast Fourier transform (FFT) algorithm for cognitive radio communication systems uses a homogeneous multiprocessor system on chip. The allows pruning of inputs such that complexity can be reduced whenever several the FFT are zero. Results show significantly reduces energy consumption compared to nonpruned version.
This work presents an update to the triangle-counting portion of subgraph isomorphism static graph challenge. is motivated by a desire understand impact CUDA unified memory on problem. First, used overlap reading large data from disk with structures in GPU memory. Second, we use hints solve multi-GPU performance scaling challenges present our last submission. Finally, improve single-GPU kernel past submission introducing work-stealing dynamic algorithm persistent threads, which makes...
This paper presents a homogeneous Multi-Processor System-on-Chip (MPSoC) as baseband signal processing engine for software defined radio applications. The implementation and parallelisation of generic OFDM system is presented taking study case the physical layer IEEE 802.11a standard. MPSoC composed nine computational nodes connected in mesh topology through hierarchical network-on-chip. Each node hosts COFFEE RISC processor element. architecture was prototyped on an ALTERA STRATIX IV FPGA...
The modern wireless standards predominantly are based on OFDM communication systems. Various mobile devices in recent times support multiple and demand efficient transceiver. Hence, a transceiver the baseband hardware needs to be scalable across standards. In an transceiver, FFT computation is one of most computationally intensive power hungry modules. Design challenging task while balancing design parameters such as speed, power, area, flexibility scalability. research work this paper...
Machine-to-machine communications has emerged to provide autonomic for a wide variety of intelligent services and applications. Among different communication technologies available connecting machines, cellular-based systems have gained more attention as backhaul networks due ubiquitous coverage mobility support. The diverse ranges service requirements well machine constraints require adopting network architectures. This paper reviews three M2M architectures integrate machines into the LTE...
Stencils are a family of widely used computational patterns that play critical role in various scientific and engineering applications. Stencil computations known to be memory-bandwidth bound, thus number different techniques algorithms optimizes memory bandwidth usage have been proposed. However, existing fall short addressing the needs large stencils, particularly more advanced stencil involving non-axis aligned grid points. To handle points, methods either use 3D caching or 2D schemes...
With the expected increase in data traffic by multiple hundred-folds coming years, future networks are likely to experience large densification of small cells, with or without detailed network planning, order capacity. The biggest challenge this brings along is very high interference. Thus there an urgent need develop strategies that could help first place avoid such interference cell clusters. In paper, we propose a novel scheme for dynamic radio resource management downlink cluster heavily...
Scanned documents (e.g., faxes) are still widely used in clinical practice and prevalent Electronic Health Records (EHR). Unlocking information scanned EHRs is critical for operation research. However, it challenging as requires converting images to texts before applying extraction technologies. Here we propose a multi-modal approach (ClinicalLayoutLM) that jointly models text extracted from Optical Character Recognition (OCR) layout/image classify into different categories lab reports CT...
High-performance distributed computing systems increasingly feature nodes that have multiple CPU sockets and GPUs. The communication bandwidth between these components is non-uniform. Furthermore, can expose different capabilities components. For communication-heavy applications, optimally using challenging essential for performance. Bespoke codes with optimized may be non-portable across run-time/software/hardware configurations, existing stencil frameworks neglect communication. This work...
With the expected increase in data traffic by multiple hundredfolds coming years, future networks will experience more and deployment of small cells with or without detailed network planning to capacity. Networks populated such dense clusters likely encounter very high interference. Thus there is an urgent need develop strategies that could help avoid interference especially for cell edge users clusters. In this paper, a decentralized coordinated scheduling (CECS) has been investigated...
Online communication platforms like Slack and Microsoft teams have become increasingly crucial for a digitized workplace to improve business efficiency growth. However, these chat can overwhelm the users with unstructured long streams of back forth discussions scattered in various places. Thus, challenging follow, leading an increased likelihood missing valuable information. Moreover, unsatisfying keyword-based search, spend significant amount time read, digest, recall information from...
In this article implementation of carrier frequency offset estimate for 20MHz LTE baseband processing is discussed. (Long Term Evolution) a wireless communication standard that makes use some innovative techniques to gain very high data rates (>100Mbps). This goal such throughput also imposes design challenges the industry and academia as in case handheld mobile devices where power budget limited. Implicitly means we need more computation energy. On other hand struggling flexible hardware...
Faxes are commonly utilized in healthcare organizations to facilitate information exchange among facilities, despite the existence of electronic health record (EHR) systems. Nevertheless, manual processing faxes is both time-consuming and prone errors. Unfortunately, critical medical conveyed through often leads delays potential harm due unavailability vital clinical data when required. This predicament further exacerbated by substantial volume received on a daily basis.
This study began with a research project, called DISCvR, conducted at the IBM-ILLINOIS Center for Cognitive Computing Systems Reseach. The goal of DISCvR was to build practical NLP based AI pipeline document understanding which will help us better understand computation patterns and requirements modern computing systems. While building such prototype, an early use case came thanks 2017 IEEE/ACM International Symposium on Microarchitecture (MICRO-50) Program Co-chairs, Drs. Hillery Hunter...
Finding the right reviewers to assess quality of conference submissions is a time consuming process for organizers. Given importance this step, various automated reviewer-paper matching solutions have been proposed alleviate burden. Prior approaches, including bag-of-words models and probabilistic topic inadequate deal with vocabulary mismatch partial overlap between paper submission reviewer's expertise. Our approach, common model, jointly topics profile while relying on abstract vectors....
Assigning qualified, unbiased and interested reviewers to paper submissions is vital for maintaining the integrity quality of academic publishing system providing valuable reviews authors. However, matching thousands with potential within a limited time daunting challenge conference program committee. Prior efforts based on topic modeling have suffered from losing specific context that help define topics in publication or submission abstract. Moreover, some cases, identified are difficult...