We analyze the interaction of explicit and implicit contracts in a model with selfish and fair principals. Fair principals are willing to honor implicit agreements, whereas selfish principals are not. Principals are privately informed about their types. We investigate a separating equilibrium in which principals reveal their type through the contract offer to the agent. If this equilibrium is p...

Automated feature discovery is a fundamental problem in machine learning. Although classical feature discovery methods do not guarantee optimal solutions in general, it has been recently noted that certain subspace learning and sparse coding problems can be solved efficiently, provided the number of features is not restricted a priori. We provide an extended characterization of this optimality ...

Theory and applications have shown that there are two important types of convergence for convex functions: pointwise convergence and convergence in a topology induced by the convergence of their epigraphs. We show that these two types of convergence are equivalent on the class of convex functions which are equi-lower semicontinuous. This turns out to be maximal classes of convex functions for w...

In this paper we continue our investigations. begun in the previous paper, of describing the solution sets of constrained extremum problems inf Au). l&l-‘(p) (*) where fand f are twice continuously differentiable functionals on a reflexive Banach space V, and r-‘(p) denotes the level set of the functional f with value p E R. Considering p as a parameter in (*) we obtain results concerning the c...

This work presents a detailed analysis and discussion of a new object tracking method using semi-supervised on-line boosting1. In order to avoid the drifting problem, which presents a challenge to adaptive tracking systems, the new approach incorporates prior knowledge of the tracked object into the tracking process via semi-supervised learning. This method makes it possible to distinguish betw...

The Problem: Many important learning problems involve a compromise between competing sources of information. For example, in semi-supervised classification the available training examples typically include a limited set of labeled data as well as a large dataset of unlabeled examples. A number of other problems share the same abstract structure: in estimating graphical models we often need to d...