Abstract:
We examine the impact of US economic news releases
in the liquidity of eleven not so extensively researched emerging stock
markets. We employ ten liquidity measures. The sample begins from June 2007 up
to December 2016. Analysis is performed in a weekly frequency. China is the
least liquid Asian market. Peru is the most liquid Latin American market. Most
of the emerging markets are positively affected by the US news, offering diversification
benefits to international investors. India
and Argentina (China and Chile) are the Asian and Latin American countries with
the highest (lowest) impacts, respectively. There
is not a single best-in-class liquidity measure. The country with the lowest
liquidity has the lowest impact from the US news releases. This result holds
for both groups of countries in Asia and Latin America.

Abstract:
We construct branched double coverings by certain direct products of manifolds for connected sums of copies of sphere bundles over the 2-sphere. As an application we answer a question of Kotschick and Loeh up to dimension five. More precisely, we show that: (1) every simply connected, closed four-manifold admits a branched double covering by a product of the circle with a connected sum of copies of $S^2 \times S^1$, followed by a collapsing map; (2) every simply connected, closed five-manifold admits a branched double covering by a product of the circle with a connected sum of copies of $S^3 \times S^1$, followed by a map whose degree is determined by the torsion of the second integral homology group of the target.

Abstract:
Every closed oriented manifold $M$ is associated with a set of integers $D(M)$, the set of self-mapping degrees of $M$. In this paper we investigate whether a product $M\times N$ admits a self-map of degree $d$, when neither $D(M)$ nor $D(N)$ contains $d$. We find sufficient conditions so that $d\notin D(M\times N)$, obtaining, in particular, products that do not admit self-maps of degree $-1$ (strongly chiral), that have finite sets of self-mapping degrees (inflexible) and that do not admit any self-map of degree $dp$ for a prime number $p$. Furthermore we obtain a characterization of odd-dimensional strongly chiral hyperbolic manifolds in terms of self-mapping degrees of their products.

Abstract:
We use work of Kotschick and Loeh to define a pure subclass of irreducible groups, called groups not infinite-index presentable by products or not IIPP. We prove that certain aspherical manifolds with fundamental groups not IIPP do not admit maps of non-zero degree from direct products. This extends the results of Kotschick and Loeh, providing new classes of aspherical manifolds - beyond those non-positively curved ones which were predicted by Gromov - that do not admit maps of non-zero degree from direct products. A sample application is that an aspherical geometric 4-manifold admits a map of non-zero degree from a direct product if and only if it is a virtual product itself. This completes a characterization of the product geometries due to Hillman. Another application is an ordering of all 4-dimensional aspherical Thurston geometries that are not hyperbolic, extending in particular an ordering of Wang in dimension three. Along the way we prove that for certain groups the property IIPP is a criterion for reducibility. This especially implies the vanishing of the simplicial volume of the corresponding aspherical manifolds. It is shown that aspherical manifolds with reducible fundamental groups do always admit maps of non-zero degree from direct products.

Abstract:
Ubiquitous automated data collection at an unprecedented scale is making available streaming, real-time information flows in a wide variety of settings, transforming both science and industry. Learning algorithms deployed in such contexts often rely on single-pass inference, where the data history is never revisited. Learning may also need to be temporally adaptive to remain up-to-date against unforeseen changes in the data generating mechanism. Online Bayesian inference remains challenged by such transient, evolving data streams. Nonparametric modeling techniques can prove particularly ill-suited, as the complexity of the model is allowed to increase with the sample size. In this work, we take steps to overcome these challenges by porting information theoretic heuristics, such as exponential forgetting and active learning, into a fully Bayesian framework. We showcase our methods by augmenting a modern non-parametric modeling framework, dynamic trees, and illustrate its performance on a number of practical examples. The end product is a powerful streaming regression and classification tool, whose performance compares favorably to the state-of-the-art.

Abstract:
The area under the ROC curve is widely used as a measure of performance of classification rules. However, it has recently been shown that the measure is fundamentally incoherent, in the sense that it treats the relative severities of misclassifications differently when different classifiers are used. To overcome this, Hand (2009) proposed the $H$ measure, which allows a given researcher to fix the distribution of relative severities to a classifier-independent setting on a given problem. This note extends the discussion, and proposes a modified standard distribution for the $H$ measure, which better matches the requirements of researchers, in particular those faced with heavily unbalanced datasets, the $Beta(\pi_1+1,\pi_0+1)$ distribution. [Preprint submitted at Pattern Recognition Letters]

Abstract:
Classical approaches for asymptotic convergence to the global average in a distributed fashion typically assume timely and reliable exchange of information between neighboring components of a given multi-component system. These assumptions are not necessarily valid in practical settings due to varying delays that might affect transmissions at different times, as well as possible changes in the underlying interconnection topology (e.g., due to component mobility). In this work, we propose protocols to overcome these limitations. We first consider a fixed interconnection topology (captured by a - possibly directed - graph) and propose a discrete-time protocol that can reach asymptotic average consensus in a distributed fashion, despite the presence of arbitrary (but bounded) delays in the communication links. The protocol requires that each component has knowledge of the number of its outgoing links (i.e., the number of components to which it sends information). We subsequently extend the protocol to also handle changes in the underlying interconnection topology and describe a variety of rather loose conditions under which the modified protocol allows the components to reach asymptotic average consensus. The proposed algorithms are illustrated via examples.

Abstract:
Consensus strategies find a variety of applications in distributed coordination and decision making in multi-agent systems. In particular, average consensus plays a key role in a number of applications and is closely associated with two classes of digraphs, weight-balanced (for continuous-time systems) and bistochastic (for discrete-time systems). A weighted digraph is called balanced if, for each node, the sum of the weights of the edges outgoing from that node is equal to the sum of the weights of the edges incoming to that node. In addition, a weight-balanced digraph is bistochastic if all weights are nonnegative and, for each node, the sum of weights of edges incoming to that node and the sum of the weights of edges out-going from that node is unity; this implies that the corresponding weight matrix is column and row stochastic (i.e., doubly stochastic). We propose two distributed algorithms: one solves the weight-balance problem and the other solves the bistochastic matrix formation problem for a distributed system whose components (nodes) can exchange information via interconnection links (edges) that form an arbitrary, possibly directed, strongly connected communication topology (digraph). Both distributed algorithms achieve their goals asymptotically and operate iteratively by having each node adapt the (nonnegative) weights on its outgoing edges based on the weights of its incoming links (i.e., based on purely local information). We also provide examples to illustrate the operation, performance, and potential advantages of the proposed algorithms.

Abstract:
Data collection at a massive scale is becoming ubiquitous in a wide variety of settings, from vast offline databases to streaming real-time information. Learning algorithms deployed in such contexts must rely on single-pass inference, where the data history is never revisited. In streaming contexts, learning must also be temporally adaptive to remain up-to-date against unforeseen changes in the data generating mechanism. Although rapidly growing, the online Bayesian inference literature remains challenged by massive data and transient, evolving data streams. Non-parametric modelling techniques can prove particularly ill-suited, as the complexity of the model is allowed to increase with the sample size. In this work, we take steps to overcome these challenges by porting standard streaming techniques, like data discarding and downweighting, into a fully Bayesian framework via the use of informative priors and active learning heuristics. We showcase our methods by augmenting a modern non-parametric modelling framework, dynamic trees, and illustrate its performance on a number of practical examples. The end product is a powerful streaming regression and classification tool, whose performance compares favourably to the state-of-the-art.

Abstract:
A key objective in systems and cognitive neuroscience is to establish associations between behavioral measures and concurrent neuronal activity. Single-trial analysis has been proposed as a novel method for characterizing such correlates by first extracting neural components that maximally discriminate trials on a categorical variable, (e.g., hard vs. easy, correct vs. incorrect etc.), and then correlate those components to a continues dependent variable of interest, e.g., reaction time, difficulty Index, etc. However, often times in experiment design it is difficult to either define meaningful categorical variables, or to record enough trials for the method to extract the discriminant components. Experiments designed for the study of the effects of stimulus presentation modality in working memory provide such a scenario, as will be exemplified. In this paper, we proposed a new approach to single-trial analysis in which we directly extract neural activity that maximally correlates to single-trial manual response times; eliminating the need to define an arbitrary categorical variable. We demonstrate our method on real electroencephalography (EEG) data recordings from the study of stimulus presentation modality effect (SPME).