In calibration mode, the tool generates several models (here, 6), and displays the properties of the models and simulation results.

1
(No Transcript)
2
Example

• In calibration mode, the tool generates several
  models (here, 6), and displays the properties of
  the models and simulation results.
• The user ranks the models and may decide to
  hand-tune some of the parameters.
• The user can also modify the weights assigned to
  the various components of the fitness function as
  well as the properties of the evolutionary
  algorithm.
• A new generation is then produced.

3
Example 2
4
Example 2
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
End
Begin
Redfast Blueslow Greenstart
9
(No Transcript)
10
(No Transcript)
11
Validation Experiment Results
Jewel - Unseen store - Medium length segment (3)
predicted
real
accuracy83.2
error
12
(No Transcript)
13
(No Transcript)
14
Geo-spatial model of mobile consumers
Densities of GSM 900, GSM 1800, UMTS and DSL
subscribers
15
Future Wi-Fi density scenarios
Sharing 50
Sharing 25
Density 100
Density 200
16
(No Transcript)
17
(No Transcript)
18
Interface
19
Tiling
20
Freezing vertical stripes size and color
21
Few lines, many circles retro
22
A sample of findings
23
Oriented search
24
Designing the GIG Enterprise Services framework
25

• Enable a better understanding of what concepts
  work and which ones could be improved upon.
• Does the computational view support the
  operational view?
• What-if Scenarios.
• User Patterns.
• Unexpected services combinations.

26
Service-chaining
WMS3
?
update
27

• Set of users
• Each has a profile
• Poisson stream of requests
• Set of servers
• Provide (sub)services
• CPU processing jobs
• Set of files
• Users servers defined
• Randomly
• By template

28
Subservice
Simple, powerful grammar to define subservices
Subservice WCS3sim(seq(6.3,WCS5),seq(WFS7,2.45))
Tree
root
sim
seq
seq
WFS7
6.3
WCS5
2.45
Current Leaves
29
Catalog Services

• Servers that offer a catalog service contain a
  link between sub-services and servers

Server0 WFS1 ? Server1 Server2 WFS2 ? Server2
WMS1 ? Server1 Server3 Server4
ServerCatalog
30
What does this all mean?
Catalog service
Catalog service
Q Who provides WFS2 ?
Q Who provides WCS6 ?
A Server 56
A Server 23
Could involve catalogs
User 56
Server 23
I need WCS6
WCS6
Heres your data...
Could be local chaining of applications
Could be non-local chaining
I need WFS2
Heres your data...
Server 56
31
(No Transcript)
32
(No Transcript)
33
Efficiency
Chain Complexity Score
34
Distinct tipping point

• b1 regression coefficient for job queue length
  vs. time (y-axis)
• Constant
• NS100
• User Submission Rate

NU50
NU350
NU
Mean job queue length (y) vs time (x)
?1gt0 ? system out of equilibrium
?10 ? system at equilibrium
35
Resilient distributed storage
36

• Goal resilient storage of multiple copies of
  data throughout network
• Objectives
• low-latency access from anywhere
• data redundancy for recovery
• de-centralized data management
• robust, low maintenance
• Philosophy let each node decide locallywhat
  data to store where and when

37
Static network analysis
38
Dynamic network analysis
39
Luxury real estate club
DomainPurchase/Use behavior ClientLuxury
multi-fractional ownership club ChallengeFind
the right parameters for business model
(membership fee, annual fees, per night fee,
property price, of properties/member mix of
target demographics) ApproachAgent-based model
of client behavior based on existing
data OutcomeIcosystem-developed predictive
dashboard for testing business model scenarios.
Discovered a business model robustly delivering
40 ARR on 10M investment.
Exclusive Resorts Private Retreats Private Escapes Portofino
Deposit 375,000 275,000 75,000 210,000
Yearly Dues 15,000 9,500 6,000 15,000
Daily Fees 0 175 75 0
Refund Percent 80 100 100 80
Member/Property Ratio 6 6 6 6
Property Loan/Value Ratio 20 32 50 40
Property Purchase Price 2,500,000 1,500,000 600,000 1,500,000
Member Cap 400 (2000) 400 400 400
40
Leading retail bank
DomainRetention, Up-Sell ClientFortune 100
Financial Services Co ChallengePredict how
customers will react to charging for an online
bill payment service (will they pay, will they
not use the service, or will they
leave?) ApproachAgent-based model of
customers OutcomeIcosystem-developed predictive
model leading to the re-definition of customer
segments and showing the profitability of
charging certain segments but not others.
41
Drug marketing with social networks
DomainAdoption/Use ClientFortune 100
Pharma ChallengeIdentify physicians that should
be targeted by sales rep in a hospital
environment, expensive drug ApproachBehavioral
model of physician prescription decision-making
and other actors playing a role in decision
(nurses, pharmacologist, etc) including network
interactions between them (surveyed). OutcomeTim
e dimension of social networks is a reliable
predictor of speed of adoption and helps
prioritize sales rep allocation and speed up
sales.
Influencer-based marketing
42
Evaluate non-physician-based marketing strategy
for schizophrenia drug
DomainRetention ClientFortune 100
Pharma ChallengeExamine alternative marketing
strategy focused on community mental health
center (CMHC) personnel rather than physicians to
enhance compliance in a complex ecosystem and
adverse side effects ApproachAgent-based model
of patients, CMHC, physicians, marketing leverage
points OutcomeAnalyzed impact of new marketing
strategy on retention and switching. Designed
clear marketing message to lift clients drug
disproportionately
43
Evaluate marketing leverage points for antiviral
drug
DomainAcquisition ClientFortune 100
Pharma ChallengeEvaluate marketing leverage
points (size of investment and size of
opportunity) in a complex patient flow context
environment that involves multiple actors,
feedback loops and time delays. ApproachAgent-ba
sed model of patient flow, including patient
behavior, GPs, specialists, tests, multiple
patient and doctor segments, marketing leverage
points OutcomeDiscovered unexpected pockets of
opportunity and made client revisit assumptions
about leverage points the model found ineffective.
44
Evaluate future value of early-stage compound
DomainAcquisition, pricing ClientFortune 100
Pharma ChallengeEvaluate NPV of early-stage
compound under wide range of possible scenarios
competitive, regulatory, marketing, market
adoption. Traditional aggregate-level, static,
linear models and assumptions built to forecast
an environment that is fundamentally dynamic and
non-linear. ApproachAgent-based model of
ecosystem, run for thousands of
scenarios. OutcomeReliable probabilistic NPV
estimates. Identification of major sources of
variance in NPV.
45
Call center simulation and analysis
ClientCall Center company ChallengeDiscover
sources of inefficiencies in operations test
decentralized operations strategies while
maintaining desired service level ApproachDetail
ed simulation of call center operations driven by
consumer demand OutcomeIdentified unexpected
key variable that drove service level
performance found low impact of decentralized
strategies
46
Redesign of clinical phases
ClientFortune 100 Pharma ChallengeRedesign
Phase I and Phase II clinical trial organization
too complex, inefficient, and difficult to
plan ApproachDetailed behavioral simulation of
decision-making and processes visualization to
identify sources of inefficiencies and
design/test a more flexible networked,
portfolio-centric clinical development
organization Outcome80 reduction in cycle time
after implementation of recommended changes.
47
(No Transcript)
48
(No Transcript)
49
Oilfield services operations
ClientFortune 100 Oilfield Services
Co. ChallengeExamine alternative scheduling
methods and organizational structures to
accommodate more business requests ApproachAgent
-based model of demand, dynamics of crew,
equipment, rigs, connecting model to
PL OutcomeAnalyzed impact of crew downsizing
identified pricing inefficiency adding 10
directly to bottom line
50
Project Supersonic Business Jet
Iteration 800 Generations
Objectives Objectives Objectives Objectives
Name Value Preference Norm Factor
Acquisition Cost 81.65 0 90
Direct Operating Cost 1.0418 0 1.5
Take-Off Gross Weight 137961 0 200000
Specific Fuel Consumption 1.159 0 1.2
Boom Loudness 84.96 0.80 88
Sideline Noise 89.84 0.20 95
Flyover Noise 82.63 0 88
Approach Speed 131.21 0 150
Fit 0.9642
Fit 0.9651
Fit 0.9615
Fit 0.9651
14393
14406
14410
14420
14393
Constraints Constraints Constraints
Name Value Constraint
Sideline Noise (dB) 89.84 lt 95
Flyover Noise (dB) 82.63 lt 88
Approach Speed (kts) 131.21 lt 150
Landing Field Length (ft) 6989 lt 9000
Take-Off Field Length (ft) 5926 lt 9000
Max Overpressure 0.7337 lt 0.95
Fuel Available 17051 gt 1000
General General General General
Name Value Min Max
of PAX 12 8 12
Manuf. ROI 6 6 12
of Vehicles 500 200 500
Design Range 3573 3500 4200
Mach 1.6597 1.6 1.8
TO Thr Der. 0.8 0.8 1
Wing Wing Wing Wing
Name Value Min Max
Location 57 45 57
AR 2.3451 2 2.5
TR 0.1689 0.05 0.3
Area 3100 2300 3100
Sweep 74 67 74
F Str-Bod Int. 0.6481 0.4 0.8
F Str-Wng Int. 0.2214 0.2 0.4
A Str-Bod Int. 0.5878 0.4 0.6
A Str-Wng Int. 0.3097 0.2 0.5
TCR - root 0.025 0.025 0.045
TCR - tip 0.0395 0.025 0.045
Twist root 1.3579 -2 2
Twist - tip 0.5683 0 5
Fuselage Fuselage Fuselage Fuselage
Name Value Min Max
Length 147.87 135 160
Cabin Loc. 36 36 41
Cabin Length 39.752 39 50
Diameter 1 2.5866 2.2 3
Diameter 2 7.349 7.2 7.6
Diameter 3 7.3851 7.2 8
Diameter 4 7.5343 7.2 7.6
Diameter 5 4.5 4.5 6.5
Diameter 6 2.3893 2.3 3.1
Engine Engine Engine Engine
Name Value Min Max
Location 107.89 100 110
OPR 24.8 22 29
TIT 3340 3300 3400
FPR 2.6357 2.6 3.2
Throttle Ratio 1.2 1.2 1.23
T/W Ratio 0.41 0.41 0.45
Empennage Empennage Empennage Empennage
Name Value Min Max
Location 87 87 97
51
Real Time Mode
52
Real Time Mode Campaign

• When an Ad Campaign is run, the background turns
  the designated color of the brand
• Here, Asience (Red) is running a campaign

53
Real Time Mode Show Friends

• Displays the network of Friends
• Highlighting communication paths

54
Real Time Mode Show Family

• Family groups are bound by rectangles

55
Real Time Mode Web

• Shows Web Sites Visited by Consumers
• Spatial distance for Web Pages should not be
  considered, as it is not factored into the code

56
(No Transcript)
57
(No Transcript)
58
(No Transcript)