The ALADDIN Center

1
The ALADDIN Center

• ALgorithm ADaptation, Dissemination and
  INtegration

2
The ALADDIN Center

• ALgorithm ADaptation, Dissemination and
  INtegration

Technology Transfer Theory
Practice
3
Domains, personel and partners covered by the
Center

• Computational Biology
• D.Durand, R. Ravi
• Celera Genomics
• Hardware Verification
• E. Clarke
• Machine Learning
• A. Blum
• Computer Security
• M. Blum, L. Blum, Rudich, Tarjan
• Intertrust

• Web Indexing
• A. Frieze, J Lafferty
• Yahoo
• Networking
• M. Harchol-Balter,
• B. Maggs
• Akamai
• Astrophysics
• R. Nichol, A. Moore
• Meshing and Comp. Geom.
• G. Blelloch, O. Ghattas, G. Miller

ALGORITHMIC CORE
4
ALADDIN Directors
Co-DIRECTORS
Guy Blelloch (PI)
Lenore Blum (co-PI)
Co-PIs
BobTarjan (Princeton)
R.Ravi
Danny Sleator
5
http//www.aladdin.cs.cmu.edu
6
Co-Directors Guy Blelloch, Lenore Blum
7
Algorithm Depot
A collection of Algorithm Resources
The algorithms depot is a community supported
source of information on algorithms. It is
designed to be accessible to a broad community,
including both researchers within the algorithms
community, as well as visitors from other areas.

It was started in the summer of 2002 as part of
the Carnegie Mellon Aladdin Center and funded by
the NSF. It is currently in the alpha phase of
development so comments are welcome. (Available
shortly.)
8
Sample PROBEs (PROBlem oriented Explorations)

• MESHING
• PRIVACY IN DATA
• Computer-Human Authentication with Applications
  to AI CAPTCHA

9
Sample PROBEs (PROBlem oriented Explorations)

• MESHING

(in conjunction with SANGRIA)
Guy Blelloch
Gary Miller
Modeling Blood Flow
10
Motivating problem 1 hemodynamic devices

• Streamliner
• left ventricular assist device
• under development at UPMC (Jim Antaki)
• Numerous advantages
• Small size
• Reliability
• Low power consumption
• Less invasive
• Magnetic bearings

• Design challenge
• Overcome tendency to shear red blood cells

11
Motivating problem(s) 2hemodynamic disease
mechanisms

• Microcirculation is complex and poorly understood
  
• Abnormal microcirculation correlated with
• cardiovascular disorders
• diabetes
• cancer
• sickle-cell anemia
• Microstructural flow models can help elucidate
  disease mechanisms

Normal and sickle cells Sickle-Cell Information
Center Emory University School of
Medicine http//www.cc.emory.edu/PEDS/SICKLE
12
Microstructural blood flow modeling
Goal Develop parallel scalable geometric and
numerical algorithms and software for simulating
flows with dynamic interfaces. Apply resulting
tools to model microstructural blood flow

• Particularly difficult due to
• large relative motion between cells
• large deformations of cellular membranes

Electron micrograph of blood flow in
12mm ateriole (Rodin, 1972)
13
Lagrangian vs. Eulerian description
Lagrangian (material) framework

• Lagrangian description of motion
• Interface representation embedded in material
  description of flow
• Interfaces are well-resolved and remain sharp
• Mesh convects and deforms with flow
• But mesh quickly becomes distorted, and dynamic
  remeshing becomes necessary
• Particularly difficult in parallel
• Eulerian description of motion
• Fixed grid
• Straightforward in parallel
• Interfaces approximately resolved through some
  other means

Eulerian (spatial) framework
14
Algorithmic framework Lagrangian flow solver
aggressive remeshing
15
Delaunay triangulation insufficient
coarsening/refinement also needed
16
Blelloch-Hardwick-Miller-Talmor Parallel Delaunay
algorithm
Median bisector (red)
Projection of points onto parabola centered on
line
Horizontal projection of points onto vertical
plane through median line

1. Compute path of Delaunay edges that bisect points
2. Continue subdividing recursively
3. Switch to sequential triangulation algorithm when
   subdomain fits on a processor

17
Application to cellular flows
Cell membrane model finitely deforming elastic
membrane
Well-resolved thin lubricating layer prevents
cellular contact
18
Sangria Research challenges

• Scaling parallel meshing up to 3D
• Scaling up to 1000s of cells
• Scaling up to large numbers of processors
• Incorporating more realistic 3D membrane models
• Experimental validation of code
• Application to problems of medical interest

19
Meshing WorkshopA Mini Roundtable for Sharing
Meshing Infrastructure,Research, and Resources
September 18, 2002 Cornell
Guy Blelloch, Gary Miller, Jonathan Shewchuk
Supported by NSF-ITR ACI 0086093
NSF-ITR Aladdin
20
Meshing WorkshopA Mini Roundtable for Sharing
Meshing Infrastructure,Research, and Resources

• Goals To share in our development effort.
• Share Code
• Share Test data
• Share interfaces

21
Ideas for transfer among our community

• Repository of test data
• Common File formats (or code to translate among
  them)
• Nearly every mesher uses its own format for
  defining input
• Common interfaces
• at the file level
• at the function level

forTransferring outside the community

• Common code repository (for finished and
  documented code)
• Documented performance
• Survey articles/books
• Work with industry to try it out and give
  feedback

22
Sample PROBEs (PROBlem oriented Explorations)

• MESHING
• PRIVACY IN DATA
• Computer-Human Authentication with Applications
  to AI

Latanya Sweeney

23
Introduction to Early Detection
Many proposals for the early detection of
biological attacks and naturally occurring
outbreaks require constantly observing the
general population through data in order to
identify behaviors that imply some number of
people are acting ill. This kind of "data
surveillance" can potentially save many lives by
alerting public health officials and criminal
investigators early of potential concerns, but
such approaches typically seek the collection of
lots of disparate information on individuals.
The data, originally collected for one purpose,
is provided to a central authority for analysis.
Thus, giving rise to serious privacy and
confidentiality concerns.
24
Continuously Observe Behaviors to Detect Onset of
Symptoms
25
Mechanical decisions typically renders data
useless
Gross overview
Sufficiently anonymous
Normal operation
Sufficiently de-identified
Unusual activity
Identifiable
Suspicious activity
Readily identifiable
Outbreak suspected
Explicitly identified
Outbreak detected
Datafly Idenifiability 0..1
Detection Status 0..1
26
Explicitly identified data generates privacy
concerns which may ultimately prohibit data
sharing
Gross overview
Sufficiently anonymous
Normal operation
Sufficiently de-identified
Unusual activity
Identifiable
Suspicious activity
Readily identifiable
Outbreak suspected
Explicitly identified
Outbreak detected
Datafly Idenifiability 0..1
Detection Status 0..1
27
Levels of identifiability matching detection
status
Gross overview
Sufficiently anonymous
Normal operation
Sufficiently de-identified
Unusual activity
Identifiable
Suspicious activity
Readily identifiable
Outbreak suspected
Explicitly identified
Outbreak detected
Datafly Idenifiability 0..1
Detection Status 0..1
28
date of birth, gender, 5-digit ZIP uniquely
identifies 87.1 of USA pop.
29
date of birth, gender, 5-digit ZIP uniquely
identifies 87.1 of USA pop.
ZIP 60623, 112,167 people, 11, not 0
insufficient above the age of 55 living there.
30
date of birth, gender, 5-digit ZIP uniquely
identifies 87.1 of USA pop.
ZIP 11794, 5418 people, primarily between 19 and
24 (4666 of 5418 or 86), only 13.
31
Linking to re-identify data
Name Address Date registered Party
affiliation Date last voted
Ethnicity Visit date Diagnosis Procedure Medicatio
n Total charge
ZIP Birth date Sex
Medical Data
Voter List
32
Data Anonymity (new field of CS)

• The study of computational solutions for
  releasing data such that the data remain
  practically useful while the identities of the
  subjects of the data are not revealed.

• PROBLEMs and CHALLENGEs
• Lots of data collected and shared
• Few attributes needed to identify a person

• Early detection critical for
• Epidemics
• Bio-terrorism

33
Data in Privacy WorkshopMarch 27-28, 2003
Carnegie Mellon
The purpose of this workshop is to bring together
researchers in Epidemiology, Law, Public Health,
Medical Informatics, Cryptography and Privacy
around the subject of collecting data for
epidemics and bio-terrorism surveillance with
scientific guarantees of privacy, anonymity and
confidentiality.
34
Sample PROBEs (PROBlem oriented Explorations)

• MESHING
• PRIVACY IN DATA
• Computer-Human Authentication with Applications
  to AI CAPTCHA

Completely Automatic Public Turing Test to Tell
Computers and Humans Apart

35
 
Aladdin Seminar
Manuel Blum
Udi Manber
36
 
Aladdin Seminar
Chat Rooms
Udi Manber
37
Word verification technology developed in
collaboration with the CAPTCHA Project at
Carnegie Mellon University.
Collaboration
38
CAPTCHA Examples
39
The ALADDIN Center

• ALgorithm ADaptation, Dissemination and
  INtegration

THE END