Bayesian Data Analysis with Mathematica
Over the last ten years, Bayesian methods have grown from a specialist niche to become a mainstream development. Bayesian data analysis is a viable alternative to frequency-based, "by the book" statistics. It provides a principled approach to solving problems without "ad-hoceries". It can deal with every-day questions, as well as with problems where other methods are inconclusive or even fail. Therefore, Bayesian data analysis has become an important research tool in this era of Big Data.
This course aims at professionals, such as data scientists and Ph.D. students, working on real data problems. It provides novel insights as well as hands-on experience.
about the course
The course covers Bayes' theorem and the use of prior probabilities, the problem with maximum likelihood and over-fitting, and regression and model selection. This perennial problem of selecting a good model, while, simultaneously, finding a good fit to the data, is treated extensively. Time permitting, some examples of classification problems, kernel methods, or neural networks are shown.
Bayesian data analysis involves considerable non-linear optimization, for which the algorithms of Mathematica are well suited. A notebook with working examples is provided as course material. Emphasis is given to practical applicability and numerical efficiency.
The course is based on the excellent book: Pattern Recognition and Machine Learning
by Christopher M. Bishop (Springer). Everyone is encouraged to have a look at this book.
A basic background in probability theory is necessary as well as proficiency in Mathematica at the level of the course Programming with Mathematica
Regular courses are held at our office in Amsterdam. Sometimes we provide the courses on site. The location is then mentioned in the agenda below.
24 March 2015, 10.00-16.30
Standard rate: euro 495+VAT
***Discounted rate: euro 95+VAT for (PhD) students
Dr. Romke Bontekoe
Dr. Romke Bontekoe got his PhD in astronomy at the University of Groningen on computer simulations of colliding galaxies. He worked in space research on the reconstruction of images from IRAS satellite data, already employing (pre-)Bayesian methods. As a long time independent consultant he has collaborated in many data analysis projects in business and in the life sciences. He wished that he had started with Mathematica earlier in his professional career.