Motivation
We're rapidly entering a world where everything can be monitored and measured. But the big problem is going to be the ability of humans to use, analyze and make sense of the data. Erik Brynjolfsson, economist and director of the MIT's Center for Digital Business
This is precisely why we will study state-of-the-art methods for pattern recognition and machine learning in this seminar. The methods to be discussed range all the way from classical techniques like logistic regresssion to recently published methods such as adaptive random forest. At the end of the seminar, you should have a rough overview over recent pattern recognition and data mining methods and know where and what to read up if you are challenged by large or high-dimensional data.
The seminar is not going to be easy, but manageable. Students not previously acquainted with this field should expect to read up at the beginning of the semester. Relevant text books and papers will be distributed at the beginning of the semester or be made available previously upon request (send email to Fred Hamprecht).
The seminar starts on April 17th, 2012 at 09:30 with an introduction to literature search, presentation techniques and the publishing process. The seminar is in English to prepare the participants for life in the English-speaking wilderness. Attendance is compulsory for at least 10 out of the 14 sessions. Any remaining time slots will be filled with presentations from members of the Heidelberg Collaboratory for Image Processing.
To register, send email to Fred Hamprecht. Welcome to the ride :-)
Contents
This seminar introduces seminal work from the field of pattern recognition, e.g. on- How to reduce the dimensionality of high-dimensional data for visualization
- How to fit models robustly in the presence of outliers
- How to cluster a data set
- Exploratory vs. confirmatory data analysis
- Statistical Hypothesis testing
- etc.