Transcript Data Science Master v3x - Institute for Computing and

Data Science Master Track
Tom Heskes and Niklas Weber
Scientific questions you will study
• What is clustering?
• What is causality?
• How can you efficiently search and rank?
• How do you build reliable models from complex data?
Why are these questions important?
To help and improve our society
iCIS data science groups
• Prof. Tom Heskes
machine learning theory and applications
• Prof. Peter Lucas
Bayesian networks and eHealth
• Dr. Elena Marchiori
complex networks and machine learning
• Prof. Theo van der Weide
information systems and retrieval
iCIS data science groups
• Prof. Wessel Kraaij
information retrieval and multimedia data analysis
• Prof. Mireille Hildebrandt
privacy and legacy aspects of data mining
• Prof. Nico Karssemeijer
computer-aided diagnosis and medical imaging
• but also: Antal van den Bosch, Bert Kappen, Lutgarde Buydens, Marcel van
Gerven, Maurits Kaptein, ...
Course outline
1st semester
2nd
semester
3rd
semester
4th semester
Track Basis
Track Basis
Track
Choice
Track Basis
Research
Seminar
Track
Choice
Research Project
CS &
Society
Master Thesis
Track
Choice
Free Choice
Track
Choice
Free Choice
External
Choice
External
choice
Track basis courses
• Mandatory, key methodological aspects
• Machine Learning in Practice (6 ec)
• Information Retrieval (6 ec)
• Bayesian Networks (6 ec)
Track choice courses
• Statistical Machine Learning (6 ec)
• Natural Computing (6 ec)
• Machine Learning (9 ec)
Theory and Tools
•
•
•
•
•
Computer aided diagnosis in medical imaging (6 ec)
Bayesian Neurocognitive Modeling (6 ec)
Bioinformatics (3 ec)
Pattern Recognition for Natural Sciences (3 ec)
Text Mining (6 ec)
Applications
•
•
•
•
Law in Cyberspace (6 ec)
Foundations of Information Systems (6 ec)
Cognition and Representation (6 ec)
Business Rules Specification and Application (3 ec)
Other aspects
Research projects
• Join one of the 7 research groups within iCIS
• Can Google Trends predict outbreaks of influenza?
Nature paper correlating Google searches to influenza outbreaks led to quite some
discussion: a fluke or actual predictive power?
• What distinguishes an excellent RTS game player from an average one? The
SkillCraft data set contains many characteristics of various players that can be
mined for actual causal relationships
• Can we discover the structure of the brain and relate this to
diseases such as Alzheimer? Time series data from neural
recordings can be analyzed to distinguish healthy from
non-healthy brains.
Master thesis projects
• Steffen Janssen developed a tool to predict productivity of software projects based
on neural networks for the Dutch tax authorities
• Thomas Janssen improved the fitting of hearing aids by machine
learning for the hearing aid company GN ReSound
• Louis Onrust studied a novel machine learning method for the extraction
of brain structure from neuroimaging data
Master thesis projects
• Niels Radstake investigated Bayesian approaches to analyze mammographic
images
• Jelle Schühmacher came up with a classifier-based method for
searching large document collections
• Tom de Ruyter works on his master thesis at Xerox in Grenoble
to improve dynamic pricing for parking in LA and other US cities
Do you want to study abroad? Or an internship?
For appointments
please mail to:
[email protected]
Room HG 00.508
But first contact your study advisor about the contents of your stay abroad!
Job perspective
• Start up your own company in data analytics, become a data analysis specialist or
consultant at a larger company, or go for a PhD
Rasa Jurgelenaite
Quantitative risk analyst
at ABN AMRO
Kristel Rösken
Business analyst
at VVV Nederland
Pavol Jancura
Software design engineer
at ASML
Laurens van de Wiel
Data scientist at FlxOne
Max Hinne and Wout Megchelenbrink
PhD students
Bart Bakker
Senior scientist
at Philips Research
Alex Slatman
Director at OBI4wan
Why Data Science at the Radboud University?
• Diversity: various aspects and applications of data science
• Flexibility: large choice of courses to shape
student interests
• Excellence: students are embedded in
research groups
Example: Machine Learning in Practice
• Basic idea: student teams enter an ongoing machine learning competition
• While trying to beat the other teams, students learn the ins and outs of challenging
machine learning problems
• Example: learn to detect whale calls in order to
prevent collisions
• The Radboud team called UHURA ended in the top
quarter of more than 200 contenders
Example: Statistical Machine Learning
• Theoretical underpinning of machine learning methods
- regression
- classification
- neural networks
- kernel methods
- mixture models and EM
• Programming and math exercises
• Demonstrations on actual data
Example: Natural Computing
• Formerly bio-inspired algorithms
• Basic idea: student teams choose a problem and solve it using bioinspired methods
• My project: use mechanisms from immune systems to develop a
method for optimization and implement this on a GPU
Example: Bayesian Neurocognitive Modeling
• Use machine learning tools to understand our brain
• Example: decode fMRI data to
reconstruct the image the person is
looking at
• Pioneered by Gallant's lab at UCB
• In the course we implement similar
techniques for still images. And that
is just one week
My impressions
•
•
•
•
•
•
Is it fun?
Is it difficult?
Can you make a living?
Will you have options? Can you reconsider?
Study environment
Should you do it?
• Pro tips:
- Have a look at some statistics before starting the courses
- Always ask. Always.