PhD Dissertation Defense

Tuesday, August 23, 2016 at 3:00 PM in Rice 242

Jian Xiang

Advisor: John Knight

Committee Members: Jack Davidson (Committee Chair), Kevin Sullivan, Hongning Wang and Houston Wood (Minor Representative).

Interpreted Formalism: Towards System Assurance and the Real-World Semantics of Software

ABSTRACT

Software systems, especially cyber-physical systems, sense and influence real-world entities under the control of software logic in order to realize desired real-world behaviors. Such software systems are based upon three essential components: (1) a computing platform, (2) a set of physical entities with which the computing platform interacts, and (3) the relationship between the first two components. These three components seem familiar, and the third component seems trivial. In fact, the third component, the relationship, is crucial, because it defines how logical values read and produced by the computing platform will be affected by and will affect the various physical entities.

Formally, the relationship between real-world entities and a computer system’s logic is the interpretation of the logic. Software logic is necessarily formal, but, in practice, interpretations are usually documented informally and incompletely, and programmers often treat elements in software logic as if they were the real-world entities themselves. As a result, faults are introduced into systems due to unrecognized discrepancies, and executions end up violating constraints inherited from the real world. The results are software and system failures and adverse downstream consequences.

This dissertation argues that, to mitigate such risks, software engineers should produce not just traditional software, but a new engineering structure, the interpreted formalism. This structure combines software logic with an explicitly documented interpretation. Among other things, an interpretation documents differences that arise inevitably between real-world values and corresponding logic values. An interpreted formalism provides centralized documentation of a system’s software and its intended relationship to the real world in an analyzable form, thereby facilitating fault detection. An implementation of the interpretation, known as real-world types, is introduced. For a specific software system, an interpretation is composed of a set of real-world types, and an interpreted formalism is implemented as a real-world type system combined with a software system.

The pragmatics of the interpreted formalism concept are illustrated by conducting case studies on two open-source software systems of different sizes. The interpreted formalism is evaluated from several viewpoints: (a) overall feasibility, (b) error detection capability, (c) effort level required, and (d) scalability. The results of the case studies suggest that: (1) the interpreted formalism concept can be used on modern software systems of different sizes, (2) the technology is capable of detecting real errors that violate real-world constraints, and (3) the effort required from engineers developing and using interpreted formalisms can be reduced greatly by an automated synthesis framework developed as part of this research.

PhD Proposal Presentation

Monday, August 8, 2016 at 2:00 PM in Rice 504

Dezhi Hong

Advisor:  Kamin Whitehouse

Committee Members:  Jack Stankovic (Committee Chair), Hongning Wang, Quanquan Gu (Systems Engineering), David Culler, (UC Berkeley).

Toward Robust and Accurate Human Activity Recognition using Wearable Sensors

ABSTRACT

Ubiquitous and immersive sensing equipment and devices, e.g., the Internet of Things, are generating an explosive amount of data. To extract meaningful and actionable information out  of these data inevitably requires the metadata of the generated data. However, the majority of  data remain unlabeled and the generation of metadata still involves labor intensive efforts, thus fundamentally unscalable. As a solution, we propose a framework for inferring the contextual information  embedded in the sensor time series, e.g., what they measure, where they locate, how they relate to  each other, etc. We take a representative of the metadata inference problem, particularly, for commercial buildings, and demonstrate first steps towards a metadata inference solution that requires minimal human intervention. At core of our solution lie a suite of techniques that exploit both the textual and time series data of sensors in buildings. We have explored a few approaches to inferring the type and location information that show promise. Building upon the early results, we  will next focus on inferring the relationship between points. This proposal provides an overview of  our solution, along with some preliminary results and key challenges, proposed research and an evaluation plan.

Master’s Project Presentation

Wednesday, July 27, 2016 at 4:00 PM in Rice 504

Jinlong Feng

Advisor: Baishakhi Ray

Committee Members: Hongning Wang

Title: Improving Spectrum-based Fault Localization using Unnaturalness of Bug

ABSTRACT: Debugging is an important but costly process during software development. Various techniques has been proposed to predict suspicious entities in the source code to assist debugging process. Such bug prediction methodologies, in general, assign suspicious scores to each program entity and generate a ranking based on the score to identify the buggy lines. For example, since buggy code has unusual behavior, they are usually considered “irregular” from the source code perspective. Natural Language Processing techniques can help evaluate this “irregularity” by building language model upon source code. On the other hand, program spectrum predicts defects by dynamic slicing—it records execution information of a program during tests, the lines caused test failures are more likely to be buggy. Both techniques generate a “suspicious” score (or ranking) for each program element, however, either technique only considered one aspect of the problem and have their own limitations. This project explores a joint analyzation of the results of these two types of software fault localization methods in order to improve the buggy line ranking results.

PhD Proposal Presentation

Thursday, July 7, 2016 at 10:00 AM in Rice 504

Md Abu Sayeed Mondol

Advisor: John Stankovic

Committee Members: Alf Weaver (Committee Chair), Jack Davidson, Yanjun Qi and John Lach (ECE, Minor Representative).

Title: Toward Robust and Accurate Human Activity Recognition using Wearable Sensors

ABSTRACT: Human activity recognition is very important in many areas including healthcare, safety, behavior monitoring, energy management and manufacturing. Wearable devices enriched with sensors like accelerometers, gyroscopes and magnetometers can be used in recognizing wide range of human activities. However, activity recognition using these devices is challenging due to issues like confounding gestures present in different activities, diversity in performing the same activity, and limited resources of the wearable devices. Today, many solutions for wearables are focused on some particular activities, and they do not generalize to other activities. One challenge is to develop underlying algorithmic solutions for activity recognition that can be used in many different wearable based applications. We propose new directions for creating such basic results that we are calling (i) Direction Agnostic Modeling, (ii) Direction Aware Modeling, (iii) Orientation Reachability, (iv) Spatiotemporal Segmentation, and (vi) Dynamic Space Time Warping for Device Orientation. Proposed techniques are based on the orientations of the wearable devices where the orientations provide pivotal information regarding different human activities. We hypothesize that highly robust and accurate activity recognition models can be developed by integrating the proposed fundamental techniques with the state of the art. The proposed research aims at finding solutions for effective integration of the techniques toward robust and accurate human activity recognition, particularly for realistic settings.

New Computer Science Faculty Members

2016 New Computer Science Faculty Members

Chang,K
Kai-Wei Chang
Assistant Professor

PhD, University of Illinois at Urbana Champaign
Post Doc, Microsoft Research, New England

Kai-Wei Chang

Research Interests: Machine Learning, Natural Language Processing and Data Mining.

WEBSITE

E-mail: kwchang@virginia.edu

ordonez-roman
Vicente Ordonez-Roman
Assistant Professor

PhD, University of North Carolina
Post Doc, The Allen Institute for AI

Vicente Ordonez-Roman

Research Interests: Computer Vision and Natural Language Understanding.

WEBSITE

E-mail: vicente@cs.virginia.edu

Shen,Helen_001
Helen Shen
PhD, Wayne State University

Helen Shen

Research Interests: Cloud computing and datacenters, Big data, Cyber-physical systems, Distributed systems, Mobile computing, High performance computing, and Social networks.

WEBSITE

2017 New Computer Science Faculty Members

Lu. FENG.photo

Lu Feng
PhD, Computer Science
Oxford, 2014

Lu Feng (January 2017)

Research Interests: Safety, security and reliability of cyber-physical systems, through formal methods and data-driven approaches, with particular emphasis on probabilistic modeling and quantitative verification.

WEBSITE

E-MAIL

Behl,Madhur
Madhur Behl
PhD, Electrical Systems Engineering,
U. of Pennsylvania, 2016

Madhur Behl (August 2017)

Research Interests: My research develops the foundations of Cyber-Physical Systems (CPS). This involves finding analytical and practical solutions to problems of modeling, control, simulation, operation and implementation of CPS. Such systems feature tightly coupled computation and communication substrates used to control large, complex, and ”messy” physical plants. CPS are the next generation of time-critical and safety-critical, networked and embedded control systems, in domains spanning from energy-efficient buildings and smart cities, to industrial automation, advanced manufacturing, autonomous vehicles and medical devices. My main research focus is centered on data-driven modeling and control of CPS, which is at the confluence of machine learning, control theory, embedded systems and statistics.

WEBSITE

E-mail: mbehl@seas.upenn.edu

Campbell,Brad
Brad Campbell
PhD candidate, U. Michigan Computer Science

Brad Campbell (August 2017)

Research Interests: My research focuses on low power hardware, wireless sensor networks, and building applications on top of the devices. Recently I’ve been looking into energy-harvesting power supplies, and their effects on hardware platforms, software stacks, and sensor networks.

WEBSITE

E-mail: bradjc@umich.edu

 Click here for:

Prof. Jim Cohoon selected for the Harold S. Morton Jr. Undergraduate Teaching Prize

cohoon-headshotCongratulations to Jim Cohoon for earning the Harold S. Morton Jr. Undergraduate Teaching Prize.  This recognizes him for his dedication to undergraduate education, his work in the development of introductory computing courses for students with little prior experience, as well as his work to promote diversity in computer science.

2015 – 2016 CS Student Awards

IMG_1816IMG_1819IMG_1807

Congratulations to Our Student Awardees

on all your hard work this year!

Graduate Student Awards

Outstanding Service – William Hawkins

Outstanding Research – Juhi Ranjan

Outstanding Teaching – Chunkun Bo and

Jinlong (Frank) Feng

Undergraduate Student Awards

Rader Award for Education

  • David Amin
  • Jeremy Gabalski
  • Stefanie Van Rafelghem
  • Nathaniel Rathjen
  • Jacqueline Tran
  • Iordan Trenkov

Rader Award for Service

  • Martin Kellogg
  • Marina Sanusi

Rader Award for Research

  • Joseph Barrow
  • Harang Ju
  • Martin Kellogg
  • Tara Raj