University of Trento

About: University of Trento is a education organization based out in Trento, Italy. It is known for research contribution in the topics: Population & Large Hadron Collider. The organization has 10527 authors who have published 30978 publications receiving 896614 citations. The organization is also known as: Universitá degli Studi di Trento & Universita degli Studi di Trento.

Hubness and Pollution: Delving into Cross-Space Mapping for Zero-Shot Learning

Abstract: Zero-shot methods in language, vision and other domains rely on a cross-space mapping function that projects vectors from the relevant feature space (e.g., visualfeature-based image representations) to a large semantic word space (induced in an unsupervised way from corpus data), where the entities of interest (e.g., objects images depict) are labeled with the words associated to the nearest neighbours of the mapped vectors. Zero-shot cross-space mapping methods hold great promise as a way to scale up annotation tasks well beyond the labels in the training data (e.g., recognizing objects that were never seen in training). However, the current performance of cross-space mapping functions is still quite low, so that the strategy is not yet usable in practical applications. In this paper, we explore some general properties, both theoretical and empirical, of the cross-space mapping function, and we build on them to propose better methods to estimate it. In this way, we attain large improvements over the state of the art, both in cross-linguistic (word translation) and cross-modal (image labeling) zero-shot experiments.

Lazy Satisfiability Modulo Theories

Abstract: Satisfiability Modulo Theories (SMT) is the problem of deciding the satisfiability of a first-order formula with respect to some decidable first-order theory T (SMT (T )). These problems are typically not handled adequately by standard automated theorem provers. SMT is being recognized as increasingly important due to its applications in many domains in different communities, in particular in formal verification. An amount of papers with novel and very efficient techniques for SMT has been published in the last years, and some very efficient SMT tools are now available. Typical SMT (T ) problems require testing the satisfiability of formulas which are Boolean combinations of atomic propositions and atomic expressions in T , so that heavy Boolean reasoning must be efficiently combined with expressive theory-specific reasoning. The dominating approach to SMT (T ), called lazy approach, is based on the integration of a SAT solver and of a decision procedure able to handle sets of atomic constraints in T (T -solver), handling respectively the Boolean and the theory-specific components of reasoning. Unfortunately, neither the problem of building an efficient SMT solver, nor even that of acquiring a comprehensive background knowledge in lazy SMT, is of simple solution. In this paper we present an extensive survey of SMT, with particular focus on the lazy approach. We survey, classify and analyze from a theory-independent perspective the most effective techniques and optimizations which are of interest for lazy SMT and which have been proposed in various communities; we discuss their relative benefits and drawbacks; we provide some guidelines about their choice and usage; we also analyze the features for SAT solvers and T-solvers which make them more suitable for an integration.' The ultimate goals of this paper are to become a source of a common background knowledge and terminology for students and researchers in different areas, to provide a reference guide for developers of SMT tools, and to stimulate the cross-fertilization of techniques and ideas among different communities.

Exosomes secreted by cortical neurons upon glutamatergic synapse activation specifically interact with neurons

Abstract: Exosomes are nano-sized vesicles of endocytic origin released into the extracellular space upon fusion of multivesicular bodies with the plasma membrane. Exosomes represent a novel mechanism of cell–cell communication allowing direct transfer of proteins, lipids and RNAs. In the nervous system, both glial and neuronal cells secrete exosomes in a way regulated by glutamate. It has been hypothesized that exosomes can be used for interneuronal communication implying that neuronal exosomes should bind to other neurons with some kind of specificity. Here, dissociated hippocampal cells were used to compare the specificity of binding of exosomes secreted by neuroblastoma cells to that of exosomes secreted by cortical neurons. We found that exosomes from neuroblastoma cells bind indiscriminately to neurons and glial cells and could be endocytosed preferentially by glial cells. In contrast, exosomes secreted from stimulated cortical neurons bound to and were endocytosed only by neurons. Thus, our results demonstrate for the first time that exosomes released upon synaptic activation do not bind to glial cells but selectively to other neurons suggesting that they can underlie a novel aspect of interneuronal communication. Keywords: extracellular vesicles; exosomes; neurons; multivesicular bodies; CD63 tetraspanin; tetanus toxin; central nervous system; intercellular communication (Published: 13 November 2014) Citation: Journal of Extracellular Vesicles 2014, 3 : 24722 - http://dx.doi.org/10.3402/jev.v3.24722

Combining Language and Vision with a Multimodal Skip-gram Model

Abstract: We extend the SKIP-GRAM model of Mikolov et al. (2013a) by taking visual information into account. Like SKIP-GRAM, our multimodal models (MMSKIP-GRAM) build vector-based word representations by learning to predict linguistic contexts in text corpora. However, for a restricted set of words, the models are also exposed to visual representations of the objects they denote (extracted from natural images), and must predict linguistic and visual features jointly. The MMSKIP-GRAM models achieve good performance on a variety of semantic benchmarks. Moreover, since they propagate visual information to all words, we use them to improve image labeling and retrieval in the zero-shot setup, where the test concepts are never seen during model training. Finally, the MMSKIP-GRAM models discover intriguing visual properties of abstract words, paving the way to realistic implementations of embodied theories of meaning.