Generalization Bounds for Learning with Linear and Quadratic Side Knowledge
نویسندگان
چکیده
In this paper, we consider a supervised learning setting where side knowledge is provided about the labels of unlabeled examples. The side knowledge has the effect of reducing the hypothesis space, leading to tighter generalization bounds, and thus possibly better generalization. We consider two types of side knowledge, the first leading to linear constraints on the hypothesis space, and the second leading to quadratic constraints on the hypothesis space. We show how different types of domain knowledge can lead directly to these kinds of side knowledge. We prove bounds on complexity measures of the hypothesis space for quadratic side knowledge, and show that these bounds are tight in a specific sense.
منابع مشابه
Robust Quadratic Assignment Problem with Uncertain Locations
We consider a generalization of the classical quadratic assignment problem, where coordinates of locations are uncertain and only upper and lower bounds are known for each coordinate. We develop a mixed integer linear programming model as a robust counterpart of the proposed uncertain model. A key challenge is that, since the uncertain model involves nonlinear objective function of the ...
متن کاملDetermining the Optimal Value Bounds of the Objective Function in Interval Quadratic Programming Problem with Unrestricted Variables in Sign
In the most real-world applications, the parameters of the problem are not well understood. This is caused the problem data to be uncertain and indicated with intervals. Interval mathematical models include interval linear programming and interval nonlinear programming problems.A model of interval nonlinear programming problems for decision making based on uncertainty is interval quadratic prog...
متن کاملA Note on Quadratic Maps for Hilbert Space Operators
In this paper, we introduce the notion of sesquilinear map on Β(H) . Based on this notion, we define the quadratic map, which is the generalization of positive linear map. With the help of this concept, we prove several well-known equality and inequality...
متن کاملChromatic PAC-Bayes Bounds for Non-IID Data: Applications to Ranking and Stationary β-Mixing Processes
PAC-Bayes bounds are among the most accurate generalization bounds for classifiers learned from independently and identically distributed (IID) data, and it is particularly so for margin classifiers: there have been recent contributions showing how practical these bounds can be either to perform model selection (Ambroladze et al., 2007) or even to directly guide the learning of linear classifie...
متن کاملSolving the Single Machine Problem with Quadratic Earliness and Tardiness Penalties
Nowadays, scheduling problems have a considerable application in production and service systems. In this paper, we consider the scheduling of n jobs on a single machine assuming no machine idleness, non-preemptive jobs and equal process times. In many of previous researches, because of the delivery dalays and holding costs, earliness and tardiness penalties emerge in the form of linear combin...
متن کامل