Data Analysis & Machine Learning

zoomzurichAI and Robotics

Oct 16, 2013 (3 years and 5 months ago)

64 views

Internet Based Communication Networks and Services • IBCN


Department of Information Technology • Ghent University

Gaston Crommenlaan 8 bus 201, 9050 Gent, Belgium

Prof. Piet Demeester

piet.demeester@intec.ugent.be

www.ibcn.intec.ugent.be


Experimental

design

is

a

structured

technique

that

helps

to

uncover

relationships

often

hidden

inside

mountains

of

data
.

The

goal

is

to

gain

knowledge

and

understanding

of

a

system

based

on

a

(limited)

set

of

well
-
chosen

experiments

over

the

design

space

of

interest
.










s

sumo
.
intec
.
ugent
.
be







Sequential

Experimental

Design


Model Order Reduction
(MOR) techniques reduce the
complexity and resolve redundancies of large
-
scale dynamical
systems, while preserving the overall
system behavior
.



















Data

Driven

Model

Order

Reduction


Surrogate models
are compact
scalable analytic models
that approximate
the multivariate input
-
output
behavior of
complex systems, based on
a limited
set of computationally
expensive
simulations or measurements.




















Surrogate

Modeling

and

Optimization


Due to the rapid development of information and communication
technologies (ICT) and the availability of high
-
throughput
experiments,
large
-
scale massive data
are accumulating in
various fields of scientific research and in society.



The
research cluster
on

Data Analysis and Machine Learning
focuses on the study and design of innovative algorithms for
efficient and accurate

characterization, analysis, modeling,
simulation and optimization

of complex systems.


Applications in various fields of science and engineering :


e
-
health, bio
-
informatics, video quality, electrical and

mechanical engineering,…

Data

Analysis

&

Machine

Learning

p
3

p
1

p
2

p
3

p
1

p
2

Problem domain

Aerodynamics

Automotive

Electronics

Metallurgy

Chemistry

Design

variables

width, frequency,

angle,
temperature, ..
.

Response

variables

lift, S
-
parameters,

pressure, stress, ...

Simulation Model


Fluent
®
, HSPICE
®
, CST
®
,

Comsol
®
, Abaqus
®
, ...

Costly

Optimization

Sensitivity

Analysis

Prototyping

CAD/CAM/CAE

Environment



Design

variables

Response

variables

Cheap

Surrogate Model /
Metamodel

Neural network, Kriging, SVM, rational function, spline,...