Computer Engineering Commons^™

Nn-X - A Hardware Accelerator For Convolutional Neural Networks, Vinayak A. Gokhale

Open Access Theses

Convolutional neural networks (ConvNets) are hierarchical models of the mammalian visual cortex. These models have been increasingly used in computer vision to perform object recognition and full scene understanding. ConvNets consist of multiple layers that contain groups of artificial neurons, which are mathematical approximations of biological neurons. A ConvNet can consist of millions of neurons and require billions of computations to produce one output. ^ Currently, giant server farms are used to process information in real time. These supercomputers require a large amount of power and a constant link to the end-user. Low powered embedded systems are not able to …

Developing Digital Media Platforms For Early Design, Sriram Karthik Badam

Open Access Theses

In recent times, mobile devices are becoming an integral part of our daily life. Software applications on these handheld devices are successfully migrating the traditional paper-based activities such as reading news, books, and even navigating through maps, onto the digital medium. While these applications allow information access anywhere and anytime, there is still a necessity for repurposing these digital media to support content/information creation especially in domains such as industrial design where paper-based activities are common. To utilize direct-touch tablets for collaborative conceptual design, we studied their affordances and iteratively developed a web-based wiki system, named skWiki. In this thesis, …

Design Of Robust Spin-Transfer Torque Magnetic Random Access Memories For Ultralow Power High Performance On-Chip Cache Applications, Xuanyao Fong

Open Access Dissertations

Spin-transfer torque magnetic random access memories (STT-MRAMs) based on magnetic tunnel junction (MTJ) has become the leading candidate for future universal memory technology due to its potential for low power, non-volatile, high speed and extremely good endurance. However, conflicting read and write requirements exist in STT-MRAM technology because the current path during read and write operations are the same. Read and write failures of STT-MRAMs are degraded further under process variations. The focus of this dissertation is to optimize the yield of STT- MRAMs under process variations by employing device-circuit-architecture co-design techniques. A devices-to-systems simulation framework was developed to evaluate …

Invariant Inferring And Monitoring In Robotic Systems, Hengle Jiang

School of Computing: Dissertations, Theses, and Student Research

System monitoring can help to detect abnormalities and avoid failures. Crafting monitors for today’s robotic systems, however, can be very difficult due to the systems’ inherent complexity and its rich operating environment.

In this work we address this challenge through an approach that automatically infers system invariants and synthesizes those invariants into monitors. This approach is inspired by existing software engineering approaches for automated invariant inference, and it is novel in that it derives invariants by observing the messages passed between system nodes and the invariants types are tailored to match the spatial, time, temporal, and architectural attributes of robotic …

Empirical Modeling Of Asynchronous Scalp Recorded And Intracranial Eeg Potentials, Komalpreet Kaur

Electrical & Computer Engineering Theses & Dissertations

A Brain-Computer Interface (BCI) is a system that allows people with severe neuromuscular disorders to communicate and control devices using their brain signals. BCIs based on scalp-recorded electroencephalography (s-EEG) have recently been demonstrated to provide a practical, long-term communication channel to severely disabled users. These BCIs use time-domain s-EEG features based on the P300 event-related potential to convey the user's intent. The performance of s-EEG-based BCIs has generally stagnated in recent years, and high day-to-day performance variability exists for some disabled users. Recently intracranial EEG (i-EEG), which is recorded from the cortical surface or the hippocampus, has been successfully used …

Querie: Collaborative Database Exploration, Magdalini Eirinaki, Suju Abraham, Neoklis Polyzotis, Naushin Shaikh

Faculty Publications

No abstract provided.

A Framework For Web Object Self-Preservation, Charles L. Cartledge

Computer Science Theses & Dissertations

We propose and develop a framework based on emergent behavior principles for the long-term preservation of digital data using the web infrastructure. We present the development of the framework called unsupervised small-world (USW) which is at the nexus of emergent behavior, graph theory, and digital preservation. The USW algorithm creates graph based structures on the Web used for preservation of web objects (WOs). Emergent behavior activities, based on Craig Reynolds’ “boids” concept, are used to preserve WOs without the need for a central archiving authority. Graph theory is extended by developing an algorithm that incrementally creates small-world graphs. Graph theory …

High Dimensional Data Set Analysis Using A Large-Scale Manifold Learning Approach, Loc Tran

Electrical & Computer Engineering Theses & Dissertations

Because of technological advances, a trend occurs for data sets increasing in size and dimensionality. Processing these large scale data sets is challenging for conventional computers due to computational limitations. A framework for nonlinear dimensionality reduction on large databases is presented that alleviates the issue of large data sets through sampling, graph construction, manifold learning, and embedding. Neighborhood selection is a key step in this framework and a potential area of improvement. The standard approach to neighborhood selection is setting a fixed neighborhood. This could be a fixed number of neighbors or a fixed neighborhood size. Each of these has …

Creating Autonomous Adaptive Agents In A Real-Time First-Person Shooter Computer Game, Di Wang, Ah-Hwee Tan

Research Collection School Of Computing and Information Systems

Games are good test-beds to evaluate AI methodologies. In recent years, there has been a vast amount of research dealing with real-time computer games other than the traditional board games or card games. This paper illustrates how we create agents by employing FALCON, a self-organizing neural network that performs reinforcement learning, to play a well-known first-person shooter computer game called Unreal Tournament. Rewards used for learning are either obtained from the game environment or estimated using the temporal difference learning scheme. In this way, the agents are able to acquire proper strategies and discover the effectiveness of different weapons without …

Influences Of Influential Users: An Empirical Study Of Music Social Network, Jing Ren, Zhiyong Cheng, Jialie Shen, Feida Zhu

Research Collection School Of Computing and Information Systems

Influential user can play a crucial role in online social networks. This paper documents an empirical study aiming at exploring the effects of influential users in the context of music social network. To achieve this goal, music diffusion graph is developed to model how information propagates over network. We also propose a heuristic method to measure users' influences. Using the real data from Last. fm, our empirical test demonstrates key effects of influential users and reveals limitations of existing influence identification/characterization schemes.

Ultimate Codes: Near-Optimal Mds Array Codes For Raid-6, Zhijie Huang, Hong Jiang, Chong Wang, Ke Zhou, Yuhong Zhao

School of Computing: Technical Reports

As modern storage systems have grown in size and complexity, RAID-6 is poised to replace RAID-5 as the dominant form of RAID architectures due to its ability to protect against double disk failures. Many excellent erasure codes specially designed for RAID-6 have emerged in recent years. However, all of them have limitations. In this paper, we present a class of near perfect erasure codes for RAID-6, called the Ultimate codes. These codes encode, update and decode either optimally or nearly optimally, regardless of what the code length is. This implies that utilizing these codes we can build highly efficient and …

S-Code: Lowest Density Mds Array Codes For Raid-6, Zhijie Huang, Hong Jiang, Ke Zhou, Yuhong Zhao, Chong Wang

School of Computing: Technical Reports

RAID, a storage architecture designed to exploit I/O parallelism and provide data reliability, has been deployed widely in computing systems as a storage building block. In large scale storage systems, in particular, RAID-6 is gradually replacing RAID-5 as the dominant form of disk arrays due to its capability of tolerating concurrent failures of any two disks. MDS (maximum distance separable) array codes are the most popular erasure codes that can be used for implementing RAID-6, since they enable optimal storage efficiency and efficient encoding and decoding algorithms. In this paper, we propose a new class of MDS array codes called …

Click-Through-Based Cross-View Learning For Image Search, Yingwei Pan, Ting Yao, Tao Mei, Houqiang Li, Chong-Wah Ngo, Yong Rui

Research Collection School Of Computing and Information Systems

One of the fundamental problems in image search is to rank image documents according to a given textual query. Existing search engines highly depend on surrounding texts for ranking images, or leverage the query-image pairs annotated by human labelers to train a series of ranking functions. However, there are two major limitations: 1) the surrounding texts are often noisy or too few to accurately describe the image content, and 2) the human annotations are resourcefully expensive and thus cannot be scaled up. We demonstrate in this paper that the above two fundamental challenges can be mitigated by jointly exploring the …

Lifetime Lexical Variation In Social Media, Lizi Liao, Jing Jiang, Ying Ding, Heyan Huang, Ee-Peng Lim

Research Collection School Of Computing and Information Systems

As the rapid growth of online social media attracts a large number of Internet users, the large volume of content generated by these users also provides us with an opportunity to study the lexical variation of people of different ages. In this paper, we present a latent variable model that jointly models the lexical content of tweets and Twitter users' ages. Our model inherently assumes that a topic has not only a word distribution but also an age distribution. We propose a Gibbs-EM algorithm to perform inference on our model. Empirical evaluation shows that our model can learn meaningful age-specific …

Idpal – A Partially-Adiabatic Energy-Efficient Logic Family: Theory And Applications To Secure Computing, Mihail T. Cutitaru

Electrical & Computer Engineering Theses & Dissertations

Low-power circuits and issues associated with them have gained a significant amount of attention in recent years due to the boom in portable electronic devices. Historically, low-power operation relied heavily on technology scaling and reduced operating voltage, however this trend has been slowing down recently due to the increased power density on chips. This dissertation introduces a new very-low power partially-adiabatic logic family called Input-Decoupled Partially-Adiabatic Logic (IDPAL) with applications in low-power circuits. Experimental results show that IDPAL reduces energy usage by 79% compared to equivalent CMOS implementations and by 25% when compared to the best adiabatic implementation. Experiments ranging …

Stochastic Analysis Of Cascading-Failure Dynamics In Power Grids, Mahshid Rahnamay-Naeini, Zhuoyao Wang, Nasir Ghani, Andrea Mammoli, Majeed M. Hayat

Electrical and Computer Engineering Faculty Research and Publications

A scalable and analytically tractable probabilistic model for the cascading failure dynamics in power grids is constructed while retaining key physical attributes and operating characteristics of the power grid. The approach is based upon extracting a reduced abstraction of large-scale power grids using a small number of aggregate state variables while modeling the system dynamics using a continuous-time Markov chain. The aggregate state variables represent critical power-grid attributes, which have been shown, from prior simulation-based and historical-data-based analysis, to strongly influence the cascading behavior. The transition rates among states are formulated in terms of certain parameters that capture grid's operating …

From The Editor-In-Chief, Ibrahim A. Baggili

Journal of Digital Forensics, Security and Law

We are proud to share with you this special edition issue of the JDFSL. This year, JDFSL partnered with both the 6th International Conference on Digital Forensics and Cyber Crime (ICDF2C) and Systematic Approaches to Digital Forensic Engineering (SADFE)–two prominent conferences in our field that were co-hosted. Fifty-three papers were submitted, and the Technical Program Committee accepted only 17 after a rigorous review process.

Anisotropy Modulations Of Femtosecond Laser Pulse Induced Periodic Surface Structures On Silicon By Adjusting Double Pulse Delay, Weina Han, Lan Jiang, Xiaowei Li, Qingsong Wang, Hao Li, Yongfeng Lu

Department of Electrical and Computer Engineering: Faculty Publications

We demonstrate that the polarization-dependent anisotropy of the laser-induced periodic surface structure (LIPSS) on silicon can be adjusted by designing a femtosecond laser pulse train (800 nm, 50 fs, 1 kHz). By varying the pulse delay from 100 to 1600 fs within a double pulse train to reduce the deposited pulse energy, which weakens the directional surface plasmon polarition (SPP)-laser energy coupling based on the initial formed ripple structure, the polarization-dependent geometrical morphology of the LIPSS evolves from a nearly isotropic circular shape to a somewhat elongated elliptical shape. Meanwhile, the controllable anisotropy of the two-dimensional scanned-line widths with different …

The Impact Of Fuzzy Requirements On Medical Device Software Development, Martin Mchugh, Abder-Rahman Ali, Fergal Mccaffery

Conference papers

Any software development project can experience difficulties with unclear or vague requirements. Unfortunately, this problem can be experience two fold in regulated environments such as the medical device software development industry. In the medical device software development industry, development organisations must contend with vague or “fuzzy” both the customer and regulatory bodies. As new requirements are introduced they can have a knock on effect on other requirements. These requirements should be analysed to determine if they are conflicting, cooperative, mutually exclusive and irrelevant. Only when the requirement is classified can a clear method be established as how to integrate that …

Story Guided Virtual Cultural Heritage Applications, Selma Rizvic

Journal of Interactive Humanities

Virtual cultural heritage applications, particularly virtual museums, nowadays include various forms of storytelling. Every object, site or artifact is better perceived and understood through the adjoining story. Interactive applications naturally request the storytelling to become interactive as well. This paper describes the concepts of interactive digital storytelling in our virtual museums and cultural heritage presentations and discusses their advantages and drawbacks recognized through user evaluation. We used digital stories not only to introduce visitors with the context and information on the objects, but also to enhance their navigation through virtual environments with purpose of learning and perceiving maximum amount of …