… domain adaptation bound for the target risk. While this bound stands in the spirit of common
domain adaptation … brings a new perspective on domain adaptation by deriving an upper …

We study the issue of PAC-Bayesian domain adaptation: We want to learn, from a source
domain, a majority vote model dedicated to a target one. Our theoretical contribution brings a …

… We provide a first PAC-Bayesian analysis for domain adaptation (DA) which arises when
the learning and test … This leads us to derive our DA-bound suitable for PAC-Bayes. …

… Domain Adaptation. We consider DA for binary classification tasks where X ⊆ Rd is the
input space of dimension d and Y ={−1, 1} is the label set. We have two different distributions …

… The three main PAC-Bayes theorems, that we present in the next section, have been
proposed by McAllester (1999); Seeger (2002); Langford (2005); and Catoni (2007). …

… In this section we reviewed the principal Pac-Bayes bounds analysis with no adaptation
and no concept of multi-view learning. In the next sections we will present the principal bounds …

… In PAC-Bayes, we also consider the risk of stochastic (Gibbs) predictors. For a distribution Q
… Flatness assumptions are not unusual in PAC-Bayes and we develop this connection next. …

… PAC-Bayes bound for the single-task setting. The bound will also serve us for the meta-learning
setting in the next sections. PAC-Bayes … a single-task PAC-Bayes theorem to bound the …

… PAC-Bayes techniques to provide an upper bound on the generalization error for new tasks
by adapting an … We demonstrate the effectiveness of this meta-adaptation approach for …

… study of domain adaptation for majority vote classifiers (from a source to a target domain).
We … [9] provide the following PAC-Bayesian generalization bound (based on the PAC-Bayes …