We construct an extension of diffusion geometry to multiple modalities through joint approximate diagonalization of Laplacian matrices. This naturally extends classical data analysis tools based on spectral geometry, such as diffusion maps and spectral clustering. We provide several synthetic and real examples of manifold learning, retrieval, and clustering demonstrating that the joint diffusion geometry frequently better captures the inherent structure of multi-modal data. We also show that many previous attempts to construct multimodal spectral clustering can be seen as particular cases of joint approximate diagonalization of the Laplacians.

This paper provides lower bounds on the convergence rate of Derivative Free Optimization (DFO) with noisy function evaluations, exposing a fundamental and unavoidable gap between the performance of algorithms with access to gradients and those with access to only function evaluations. However, there are situations in which DFO is unavoidable, and for such situations we propose a new DFO algorithm that is proved to be near optimal for the class of strongly convex objective functions. A distinctive feature of the algorithm is that it uses only Boolean-valued function comparisons, rather than function evaluations. This makes the algorithm useful in an even wider range of applications, such as optimization based on paired comparisons from human subjects, for example. We also show that regardless of whether DFO is based on noisy function evaluations or Boolean-valued function comparisons, the convergence rate is the same.

We show that the herding procedure of Welling (2009) takes exactly the form of a standard convex optimization algorithm--namely a conditional gradient algorithm minimizing a quadratic moment discrepancy. This link enables us to invoke convergence results from convex optimization and to consider faster alternatives for the task of approximating integrals in a reproducing kernel Hilbert space. We study the behavior of the different variants through numerical simulations. The experiments indicate that while we can improve over herding on the task of approximating integrals, the original herding algorithm tends to approach more often the maximum entropy distribution, shedding more light on the learning bias behind herding.

The aim of this paper is to develop a methodology that is useful for analysing from a microeconomic perspective the incentives to entry, permanence and exit in the market for pharmaceutical generics under fuzzy conditions. In an empirical application of our proposed methodology, the potential towards permanence of labs with different characteristics has been estimated. The case we deal with is set in an open market where global players diversify into different national markets of pharmaceutical generics. Risk issues are significantly important in deterring decision makers from expanding in the generic pharmaceutical business. However, not all players are affected in the same way and/or to the same extent. Small, non-diversified generics labs are in the worse position. We have highlighted that the expected NPV and the number of generics in the portfolio of a pharmaceutical lab are important variables, but that it is also important to consider the degree of diversification. Labs with a higher potential for diversification across markets have an advantage over smaller labs. We have described a fuzzy decision support system based on the Mamdani model in order to determine the incentives for a laboratory to remain in the market both when it is stable and when it is growing.

In this article, I consider the effects of networking on the emergence of intelligence in individuals and societies. The following hypothesis promotes and sustains this investigation: The General Collective Problem Solving Capacity Hypothesis. Society possesses a general, collective problem solving capacity. The General Collective Problem Solving Capacity Hypothesis implies that the same general problem solving capacity that society uses, for example, to develop language, is used to solve problems in mathematics, science, business, musical composition and performance, sports contests, social interactions, politics and daily life. "All life is problem solving" [47]; all problem solving is a strictly analogous process. Let's adopt some notational conventions that will allow us to make the observations in the discussion that follows more precise. The formulas used in the definitions are sometimes modified by a subscript relevant to the context in which they are used.

We investigate a class of first-order temporal-epistemic logics for reasoning about multi-agent systems. We encode typical properties of systems including perfect recall, synchronicity, no learning, and having a unique initial state in terms of variants of quantified interpreted systems, a first-order extension of interpreted systems. We identify several monodic fragments of first-order temporal-epistemic logic and show their completeness with respect to their corresponding classes of quantified interpreted systems.

Latent feature models are widely used to decompose data into a small number of components. Bayesian nonparametric variants of these models, which use the Indian buffet process (IBP) as a prior over latent features, allow the number of features to be determined from the data. We present a generalization of the IBP, the distance dependent Indian buffet process (dd-IBP), for modeling non-exchangeable data. It relies on distances defined between data points, biasing nearby data to share more features. The choice of distance measure allows for many kinds of dependencies, including temporal and spatial. Further, the original IBP is a special case of the dd-IBP. In this paper, we develop the dd-IBP and theoretically characterize its feature-sharing properties. We derive a Markov chain Monte Carlo sampler for a linear Gaussian model with a dd-IBP prior and study its performance on several non-exchangeable data sets.

We offer a novel view of AdaBoost in a statistical setting. We propose a Bayesian model for binary classification in which label noise is modeled hierarchically. Using variational inference to optimize a dynamic evidence lower bound, we derive a new boosting-like algorithm called VIBoost. We show its close connections to AdaBoost and give experimental results from four datasets.

In this article, we address the question of how non-knowledge about future events that influence economic agents' decisions in choice settings has been formally represented in economic theory up to date. To position our discussion within the ongoing debate on uncertainty, we provide a brief review of historical developments in economic theory and decision theory on the description of economic agents' choice behaviour under conditions of uncertainty, understood as either (i) ambiguity, or (ii) unawareness. Accordingly, we identify and discuss two approaches to the formalisation of non-knowledge: one based on decision-making in the context of a state space representing the exogenous world, as in Savage's axiomatisation and some successor concepts (ambiguity as situations with unknown probabilities), and one based on decision-making over a set of menus of potential future opportunities, providing the possibility of derivation of agents' subjective state spaces (unawareness as situation with imperfect subjective knowledge of all future events possible). We also discuss impeding challenges of the formalisation of non-knowledge.

The matrices and their sub-blocks are introduced into the study of determining various extensions in the sense of Dung's theory of argumentation frameworks. It is showed that each argumentation framework has its matrix representations, and the core semantics defined by Dung can be characterized by specific sub-blocks of the matrix. Furthermore, the elementary permutations of a matrix are employed by which an efficient matrix approach for finding out all extensions under a given semantics is obtained. Different from several established approaches, such as the graph labelling algorithm, Constraint Satisfaction Problem algorithm, the matrix approach not only put the mathematic idea into the investigation for finding out various extensions, but also completely achieve the goal to compute all the extensions needed.