Columbia University

1

• Columbia University
• Department of Computer Science
• Henning Schulzrinne, Chair
• Department of Computer Science

IBM CAS
2
Columbia Computer Science in Numbers

• 34 full-time faculty and lecturers
• visitors, postdocs, adjunct faculty, joint
  appointments (EE, IEOR),
• 125 PhD students (10 new arrivals)
• 221 MS students (120 new arrivals)
• 100 CS undergraduate majors (juniors, seniors)
• 20 computer engineering students

3
Faculty 34 (31 tenure track, 3 lecturers) 3
joint




Carloni
Feiner
Edwards
Aho
Allen
Belhumeur
Cannon
Galil
Bellovin
Gravano
Gross
Grunschlag
Kender
Kaiser
Keromytis
Hirschberg
Jebara
Grinspun
McKeown
Nayar
Ramamoorthi
Ross
Nowick
Nieh
Malkin
Rubenstein
Misra
Yannakakis
Servedio
Schulzrinne
Stolfo
Shortliffe
Unger
Stein
Wozniakowski
Traub
Yemini
4
Columbia Computer Science Research
UI, NLP, collab work
graphics, robotics, vision
networks, security, OS, software eng
quantum computing, crypto, learning, algorithms
databases, data mining, machine learning
CAD, async circuits, embedded systems
5
Research areas
Interacting with the Physical World graphics, robotics, vision Allen, Belhumeur, Feiner, Grinspun, Grunschlag, Jebara, Kender, Nayar, Ramamoorthi
Interacting with Humans user interfaces, natural language and speech processing, collaborative work, personalized agents Feiner, Hirschberg, Kaiser, Kender, McKeown
Systems networks, distributed systems, security, compilers, software engineering, programming languages, OS Aho, Bellovin, Edwards, Kaiser, Keromytis, Malkin, Misra, Nieh, Schulzrinne, Stolfo, Yemini
Designing Digital Systems digital and VLSI design, CAD, asynchronous circuits, embedded systems Carloni, Edwards, Nowick, Unger
Making Sense of Data databases, data mining, Web search, machine learning applications Cannon, Gravano, Jebara, Kaiser, Ross, Servedio, Stolfo
Computer Science Theory cryptography, quantum computing, complexity, machine learning theory, graph theory, algorithms Aho, Galil, Gross, Malkin, Servedio, Traub, Wozniakowski, Yannakakis
6
CCLS A Research Center in CS

• The Center for Computational Learning Systems
  (CCLS) aims to be a world leader in learning and
  data mining research and the application of this
  research to natural language understanding, the
  World Wide Web, bioinformatics, systems security
  and other emerging areas. CCLS will emphasize
  interdisciplinary efforts with other departments
  at Columbia, and will leverage Columbia's CS
  Department's strengths in learning, data mining
  and natural language processing, extending the
  effective size and scope of the Department's
  research effort.

7
Interacting with Humans Newsblaster

• Automatic summarization of articles on the same
  event
• Generation of summary sentences
• Tracking events across days
• Foreign news ? English summaries

Faculty Kathy McKeown
8
Working with IBM Text-to-Speech Synth. Group

• Joint work between Michael Pichenys TTS group
  and Julia Hirschberg and students at CU
• Issue IBMs Research TTS system one of the
  best, but even the best TTS systems often do not
  sound like humans
• intonation
• ability to convey human emotion to a listener
• Our joint goal to enhance the IBM system to
  improve naturalness and expressiveness via
• Better assignment of intonational prominence and
  phrasing
• Additional flexibility to produce emotional
  speech (certainty/uncertainty, anger/frustration)

9
Interacting with Humans Detecting Deceptive
Speech

• Problem
• Can we detect deception from spoken language cues
  only?
• Method
• Collect corpus of deceptive non-deceptive
  speech
• Extract acoustic, prosodic and lexical features
  automatically
• E.g., disfluencies, response latency, high pitch
  range, lower intensity, laughter, personal
  pronouns
• Run machine learning experiments to create
  automatic prediction models and test on held-out
  data
• Results
• Baselines
• Best general human performance in literature
  ranges from criminals (65 accuracy) down to
  parole officers (40)
• Majority class, our data (predict truth) 61
• Mean human performance with our data 60
• Our (automatic) results 69

Faculty Julia Hirschberg
10
Interacting with Humans Learning to Match Authors
Entity Resolution of Anonymized Publications 7
Teams UMass, Maryland, Fair-Isaac, Illinois,
Rutgers, CMU, Columbia
Key 1 - Permutational Text Kernels 2 -
Permutational Clustering 3 - SVM
Source 2005 KDD Challenge
Faculty Tony Jebara
11
Systems Distributed Channel Allocation in Mobile
Mesh Networks

• Multi-radio mesh node
• Channel scarcity ? need automated channel
  allocation in 802.11 mesh networks
• Allocates radios by self-stabilizing algorithm
  based on graph coloring

• Results
• First self-organizing mechanism implementation
• Network self-organizes in seconds
• Network throughput improvement of 20-100 cf.
  static channel allocation

Collaborators Victor Bahl and Jitendra Padhye _at_
MSR
Faculty Misra/Rubenstein
12
Systems Creating new services for VoIP

• Old telecom model
• Programmers create mass-market applications
• new service each decade
• Our (web) model
• Users and administrators create universe of
  tailored applications
• Incorporate human context
• location, mood, actions,
• Eclipse for service creation
• Based on presence, location, privacy preferences
• Learn based on user actions

Faculty Henning Schulzrinne
13
Self-healing Software Systems

• Novel techniques for software that repairs its
  failures based on Observe-Orient-Decide-Act
  (OODA) loop
• Demonstrated concept with two experimental
  prototypes
• One aimed at the problem of worms
• One aimed at software survivability in general
• Application Communities enable large numbers of
  identical applications to collaboratively monitor
  their health and share alerts

14
Self-patching Architecture

• Systems approach to creating software that
• Detects new attacks/failures
• Automatically generates and applies appropriate
  fixes
• Developed error virtualization as a generic
  band-aid technique
• Prototypes for open-source and binary-only
  environments
• Efficient security and high availability
  mechanism with little performance penalty
• Spin-off Revive Systems Inc.

15
Computer Graphics and User Interfaces Lab (Prof.
S. Feiner) Cursorless Input for Wearable
UIsGábor Blaskó

• Minimize need for visual feedback in wearable
  user interfaces
• E.g., watch bezel serves as tactile landmarks
  to guide users finger

16
Columbia/FSTC Relationship

• A group of 6 CS faculty manage the Security
  Standing Committee
• The CS department hosts the FSTC executive
  director, and provides facilities for meetings
• Members include most of the largest banks and
  financial institutions and IT security vendors
• Industry collaborative RD Projects dealing with
  security of the IT infrastructure

17
Conclusion

• Broad-based research motivated by real problems
• Breaking new ground in several key areas, e.g.
• Natural language processing
• New network services and models
• Network security
• Graphics vision
• Columbia has a growing impact on computer science
  as demonstrated in successfully bringing new
  technology to the field
• Start-ups
• Standardization
• Education