In particular, this means that smaller entries of A lead to random variables with smaller variance. On the other hand, the bound on ∥∥∥A− Â∥∥∥ 2 depends on the maximum variance. Thus, to improve the results, one idea is to keep entries of Aij with probability pij ≤ p, so that all entries in  have roughly the same variance. This will help us to get sparser matrices, while keeping similar qualit...

We survey two basic techniques for showing that the monadic second-order theory of a structure is decidable. In the first approach, one deals with finite fragments of the theory (given for example by the restriction to formulas of a certain quantifier rank) and – depending on the fragment – reduces the model under consideration to a simpler one. In the second approach, one applies a global tran...

In this paper, we present a framework for fitting multivariate Hawkes processes for large-scale problems, both in the number of events in the observed history n and the number of event types d (i.e. dimensions). The proposed Scalable LowRank Hawkes Process (SLRHP) framework introduces a lowrank approximation of the kernel matrix that allows to perform the nonparametric learning of the d trigger...

This paper describes our participation in the Federated Web Search track at TREC 2014. Our main focus is on the resource selection task, where we employ a learning-to-rank approach to combine various (instantiations of) resource ranking models. Further, we show that vertical selection can be run on the output from resource selection, and that it directly benefits from the improvements of thereof.

We introduce a generalization of CM-triviality relative to a fixed invariant collection of partial types, in analogy to the Canonical Base Property defined by Pillay, Ziegler and Chatzidakis which generalizes one-basedness. We show that, under this condition, a stable field is internal to the family, and a group of finite Lascar rank has a normal nilpotent subgroup such that the quotient is alm...

Let A be the coordinate ring of an affine elliptic curve (over an infinite field k) of the form X − {p}, where X is projective and p is a closed point on X. Denote by F the function field of X. We show that the image of H•(GL2(A), Z) in H•(GL2(F ), Z) coincides with the image of H•(GL2(k), Z). As a consequence, we obtain numerous results about the K-theory of A and X. For example, if k is a num...

We calculate the rational toral ranks of two spaces whose cohomologies are isomorphic and note that rational toral rank is a rational homotopy invariant but not a cohomology invariant.

This paper proposes the adoption of different word-level scores in the framework of automatic pronunciation scoring using learning to rank. Six types of phone-level scores are first computed and converted to word-level scores by using average-based, vowel-based, and consonant-based methods. Different score combination methods are then used to combine these word-level scores to obtain the final ...