Defense Notices

The Treatise on Invertebrate Paleontology (TIP) is a definitive work completed by more than 300 authors in the field of Paleontology, covering all categories of invertebrate animals. The digital version for TIP is consisted of multiple PDF files, however, these files are just a clone of paper version and are not well formatted, which makes it hard to extract structured data using only straightforward methods. In order to make fossil and extant records in TIP organized and searchable from a web interface, a digital library which is called Invertebrate Paleontology Knowledgebase (IPKB) was built for information sharing and querying in TIP. It is consisted of a database which stores records of all fossils and extant invertebrate animals, and a web interface which provides an online access. 
The existing IPKB system provides a general framework for TIP information showing and searching, however, it has very limited search functions, only allowing users querying by pure text. Details of structural properties in the fossil descriptions are not carefully taken into consideration. Moreover, sometimes users cannot provide correct and rich enough query terms. Although authors of TIP are all paleontologists, the expected users of IPKB may not be that professional. 
In order to overcome this limitation and bring more powerful search features into the IPKB system, in this thesis, we present a content-based search function, which allow users to search using textual ontology descriptions and images of fossils. First, this thesis describes the work done by previous research on IPKB system. Except for the original text and image processing approaches, we also present our new efforts on improving these original methods. Second, this thesis presents the algorithm and approach adopted in the construction of content-based search system for IPKB. The search functions in the old IPKB system did not consider the differences among morphological details of certain regions of fossils. Three major parts are discussed in detail: (1) Textual ontology based search. (2) Image based search. (3) Text-image based search. 

Participation in social networking sites has dramatically increased in recent years. Services such as Linkedin, Facebook, or Twitter allow millions of individuals to create online profiles and share personal information with vast networks of friends - and, often, unknown numbers of strangers. The relation between privacy and a person’s social network is multi-faced. At certain occasions we want information about ourselves to be know only to a limited set of people, and not to strangers. Privacy implications associated with online social networking depend on the level of identifiability of the information provided, its possible recipients, and its possible uses. Even social networking websites that do not openly expose their users’ identities may provide enough information to identify profile’s owner. 

With the advent of general-purpose programming tools and newer GPUs, programmers now have access to a more flexible general-purpose approach to using GPUs for something other than graphics. With single instruction stream, multiple data streams (SIMD), the same instruction is executed by multiple processors using different data streams. GPUs are SIMD computers that exploit data-level parallelism by applying the same operations to multiple items of data in parallel. There are many areas where GPUs can be used for general-purpose computing. We have chosen to focus on a project in the astrophysics area of scientific computing called N-body simulation which computes the evolution of a system of bodies that interact with each other. Each body represents an object such as a planet or a star, and each exerts a gravitational force on all the others. It is performed by using a numerical integration method to compute the interactions among the system of bodies, and begins with the initial conditions of the system which are the masses and starting position and velocity of every body. These data are repeatedly used to compute the gravitational force between all bodies of the system to show updates on screen. We investigate alternative implementation approaches to the problem in an attempt to determine the factors that maximize its performance, including speed and accuracy. Specifically, we compare an OpenCL approach to one based on using OpenGL Compute Shaders. We select these two for comparison to generate real-time interactive displays with OpenGL. Ultimately, we anticipate our results will be generalizable to other APIs (e.g., CUDA) as well as to applications other than the N-Body problem. A comparison of various numerical integration and memory optimization techniques is also included in our analysis in an attempt to understand how they work in the SIMD GPGPU environment and how they contribute to our performance metrics. We conclude that, for our particular implementation of the problem, taking advantage of efficiently using local memory considerably increases performance.

The number of vehicle manufacturers and the number of vehicles produced have been significantly increasing each year. With more vehicles on road, the number of accidents on the National Highways in the US increased notably. NHTSA (National Highway Traffic Safety Agency) is a federal agency which works towards preventing vehicle crashes and their attendant costs. They plan and execute several operations and control measures to find and solve the problems causing accidents. One such initiative is to analyze the primary causes of all the vehicle crashes and maintain a streamlined data of vehicle Identification catalog customized for DOT and NHTSA. Maintaining a data on about 250+ millions of vehicles and analyze them needs a robust database and an application for its maintenance. At StrongBridge Corporation, we developed VPICLIST, an application for NHTSA to assist their analytic projects with data entry and pattern decoding of VIN information catalog. The application employs precise pattern matching techniques to dump data into distributed databases which in turn collaborate to a central database of NHTSA. It allows decoding of VIN each at a time by the public and also decoding thousands of VINS simultaneously for internal use of NHTSA. To hold and operate upon several PBs of data, insertion and retrieval process of the application emulates a distributed architecture. The application is developed in Java and uses Oracle enterprise database for distributed small collections and NoSQL system for the central database.

The study of shear wave velocity (Vs) of near surface materials has been one of the primary areas of interest in seismic analyses. ‘Vs’ serves as the best indicator in evaluating the stiffness of a material from its shear modulus. One of the economical methods to obtain Vs profiling information is through the analysis of dispersion property of surface waves. SurfSeis4 - Software developed by the Kansas Geological Survey (KGS) utilizes Multichannel Analysis of Surface Waves (MASW) method to obtain shear wave velocity 2D (Surface location and depth) profiling. The profiling information is obtained in the form of a grid through inversion of dispersion curves. The Vs 2D map module of SurfSeis4, integrates the functionality of interpolating this grid to approximate the variation of shear wave velocity across the surface locations. The current project is an extension of the existing SurfSeis4 Vs 2D mapping module in its latest release of SurfSeis5 that incorporates topography in shear wave velocity variation and facilitates users with advanced image interpolation options.

One of the main components in switching power supply is switch. However, there are two main negative issues the switch will cause in a switching power supply. The first one is that the power dissipation of the switch will be unimaginable high, especially when the current go through the switch gets higher. Secondly, because there are so many parasitic inductances and capacitances in the circuit, transient will cause problems when the operating state of the switch changes. In this project, P-Spice is used to design a qualify swith and suppress the negative effect as much as possible. The purpose of this project is to design a switch for hardware design in switching power supplies. Therefore, all the components used in P-Spice simulation are the actual models which is able to get from electronic market, and all the situations which may be happen in hardware design will be consider in the simulation. Both Mosfet and bipolar transistor switch will be discussed in the project. The project will give solutions for reducing the power dissipation cause by the switch and transient problems.

Packet Filtering is a process of filtering the packets based on the filters rules that are being defines by the user. The focus of this project is to implement and compare the performance of two different packet filtering techniques (SFSA and PFSA), that uses FSA(finite state automaton) for the filtering process. Stateless FSA(SFSA) is a packet filtering technique where an FSA is generated based on the input packet and the filtering criteria. Then succeed early algorithm is applied to the automaton which simplifies by the automaton by shortening long trails to the the final state which reduces the packet filtering time. It also uses transition compaction algorithm which helps in avoiding certain areas in packet inspection which are not necessary for packet filtering. 
PFSA (predicates of FSA) does the filtering based on predicates generated by the predicate evaluator. In this filtering process the FSA generated as state transitions which depend on the input symbol and also the predicate value. In order to simplify the FSA algorithms like predicates Cartesian product and predicates anticipation algorithms are being used. These algorithms consider all states that are possible and merge them to make the FSA deterministic. There is also a proto FSA that is being generated for the predicates to speed up the filtering process. 

Neural Networks were studied and applied to many research problems from a long time. With gaining popularity of deep neural networks in the area of machine learning, many researchers in various domains want to try deep learning framework. Deep learning requires lot of memory and high processing power. One way of doing it faster is to make use of GPUs which use distributed and parallel processing, thereby increasing speed. But because of the computation (lot of vector and matrix operations) deep learning requires, expensive infrastructure required (GPUs and clusters), hardware and software installation overhead, not many researchers prefer deep learning. The current application is a solution to cheminformatics problems using Convolutional Architecture for Fast Feature Embedding (Caffe) deep learning framework. The application provides a framework/service to researchers who want to try deep learning on their datasets. The application accepts datasets from users along with options for hyper parameter configuration, runs cross fold validation on the training dataset, and makes predictions on the test dataset. The (tuning) results of running caffe on the training dataset and predictions made on test dataset are sent to user via an email. The current version supports binary classification that predicts activity/inactivity of a chemical compound based on molecular fingerprints which are binary features.

Given the high complexity and increasing traffic load of the current Internet, the geographically-correlated challenge caused by large-scale disasters or malicious attacks pose a significant threat to dependable network communications. To understand its characteristics, we start our research by first proposing a critical-region identification mechanism. Furthermore, the identified regions are incorporated into a new graph resilience metric, compensated Total Geographical Graph Diversity (cTGGD), which is capable of characterizing and differentiating resiliency levels for different topologies. We further propose the path geodiverse problem (PGD) that requires the calculation of a number of geographically disjoint paths, and two heuristics with less complexity compared to the optimal algorithm. We present two flow-diverse multi-commodity flow problems, a linear minimum-cost and a nonlinear delay-skew optimization problem to study the tradeoff among cost, end-to-end delay, and traffic skew on different geodiverse paths. We further prototype and integrate the solution from above models into our cross-layer resilient protocol stack, ResTP--GeoDivRP. Our protocol stack is implemented in the network simulator ns-3 and emulated in the KanREN testbed. By providing multiple geodiverse paths, our protocol stack provides better path protection than Multipath TCP (MPTCP) against geographically-correlated challenges. Finally, we analyze the mechanism attackers could utilize to maximize the attack impact and demonstrate the effectiveness of a network restoration plan. 

In a world with enormously growing technologies and applications, most people use smart 
devices. This provides a means to develop smart applications that will be help students learn effectively. 
In this project, we develop a smart android application which will provide digital means of 
interaction between the professors and students. Instead of using traditional emails for every 
discussion, this application helps to broadcast multiple messages to the class through a single 
click. The students will also be able to follow multiple professors and participate in the active 
discussions. And also this application allows the users to send personal messages to the other 
users in order to participate in an active discussion. It provides unique logins to every student 
and professor. It uses mongoDB as the database and "parse" backend as a service.The main 
inspiration for this project was an application called Tophat.