Suboptimal behaviour of Bayes and MDL in classification under misspecification

Suboptimal Behavior of Bayes and MDL in

Learning from Dependent Observations

09:00 – 09:55 Registration and coffee-break 09:55 – 10:00 Opening remarks 10:00 – 10:50 Emmanuel Candes (California Institute of Technology) The Dantzig Selector: Statistical Estimation when p is Larger than n 11:00 – 11:30 Coffee-break 11:30 – 12:20 Franck Barthe (Université Toulouse III) About Talagrand’s Concentration Inequality for Exponential Measures 12:30 – 14:30 Lunch break 14:30 – 15:2...

Safe Learning: bridging the gap between Bayes, MDL and statistical learning theory via empirical convexity

We extend Bayesian MAP and Minimum Description Length (MDL) learning by testing whether the data can be substantially more compressed by a mixture of the MDL/MAP distribution with another element of the model, and adjusting the learning rate if this is the case. While standard Bayes and MDL can fail to converge if the model is wrong, the resulting “safe” estimator continues to achieve good rate...

Comparison of Decision Tree and Naïve Bayes Methods in Classification of Researcher’s Cognitive Styles in Academic Environment

In today world of internet, it is important to feedback the users based on what they demand. Moreover, one of the important tasks in data mining is classification. Today, there are several classification techniques in order to solve the classification problems like Genetic Algorithm, Decision Tree, Bayesian and others. In this article, it is attempted to classify researchers to “Expert” and “No...

An analysis of the difference of code lengths between two-step codes based on MDL principle and Bayes codes

In this paper, we discuss the difference in code lengths between the code based on the minimum description length (MDL) principle (the MDL code) and the Bayes code under the condition that the same prior distribution is assumed for both codes. It is proved that the code length of the Bayes code is smaller than that of the MDL code by (1) or (1) for the discrete model class and by (1) for the pa...