Infinite Dirichlet mixture models learning via expectation propagation

In this article, we propose a novel Bayesian nonparametric clustering algorithm based on a Dirichlet process mixture of Dirichlet distributions which have been shown to be very flexible for modeling proportional data. The idea is to let the number of mixture components increases as new data to cluster arrive in such a manner that the model selection problem (i.e. determination of the number of clusters) can be answered without recourse to classic selection criteria. Thus, the proposed model can be considered as an infinite Dirichlet mixture model. An expectation propagation inference framework is developed to learn this model by obtaining a full posterior distribution on its parameters. Within this learning framework, the model complexity and all the involved parameters are evaluated simultaneously. To show the practical relevance and efficiency of our model, we perform a detailed analysis using extensive simulations based on both synthetic and real data. In particular, real data are generated from three challenging applications namely images categorization, anomaly intrusion detection and videos summarization.     

扭型Kazhdan-Lusztig理论中的逆反多项式

本文研究扭型Kazhdan-Lusztig多项式的逆反多项式的性质及其计算方法.构造了Lusztig对偶模M的一类特异基(或D-基),获得了Hecke代数在此基上的作用公式.在有限Coxeter群情形下,获得了Lusztig-Vogan模的结构常数的关系.     

An effective architecture for learning and evolving flexible job-shop schedules

In recent years, the interaction between evolution and learning has received much attention from the research community. Some recent studies on machine learning have shown that it can significantly improve the efficiency of problem solving when using evolutionary algorithms. This paper proposes an architecture for learning and evolving of Flexible Job-Shop schedules called LEarnable Genetic Architecture (LEGA). LEGA provides an effective integration between evolution and learning within a random search process. Unlike the canonical evolution algorithm, where random elitist selection and mutational genetics are assumed; through LEGA, the knowledge extracted from previous generation by its schemata learning module is used to influence the diversity and quality of offsprings. In addition, the architecture specifies a population generator module that generates the initial population of schedules and also trains the schemata learning module. A large range of benchmark data taken from literature and some generated by ourselves are used to analyze the efficacy of LEGA. Experimental results indicate that an instantiation of LEGA called GENACE outperforms current approaches using canonical EAs in computational time and quality of schedules.     

Accelerating autonomous learning by using heuristic selection of actions

This paper investigates how to make improved action selection for online policy learning in robotic scenarios using reinforcement learning (RL) algorithms. Since finding control policies using any RL algorithm can be very time consuming, we propose to combine RL algorithms with heuristic functions for selecting promising actions during the learning process. With this aim, we investigate the use of heuristics for increasing the rate of convergence of RL algorithms and contribute with a new learning algorithm, Heuristically Accelerated Q-learning (HAQL), which incorporates heuristics for action selection to the Q-Learning algorithm. Experimental results on robot navigation show that the use of even very simple heuristic functions results in significant performance enhancement of the learning rate.     

HOPF代数的右伴随作用构造的上循环模(英文)

本文研究了上循环模,对于特征为O的域k上满足S~2=id_H的Hopf代数H,和左H-模代数A,利用日的右伴随作用以及H在A上的模作用,构造了上循环模(C)_H~#(A),并且证明了由H的右伴随作用和左伴随作用分别诱导的上循环模(C)_H~(#)(A)和(C)_H~(#)(A)足同构的.     

Imprecise probability models for learning multinomial distributions from data. Applications to learning credal networks

This paper considers the problem of learning multinomial distributions from a sample of independent observations. The Bayesian approach usually assumes a prior Dirichlet distribution about the probabilities of the different possible values. However, there is no consensus on the parameters of this Dirichlet distribution. Here, it will be shown that this is not a simple problem, providing examples in which different selection criteria are reasonable. To solve it the Imprecise Dirichlet Model (IDM) was introduced. But this model has important drawbacks, as the problems associated to learning from indirect observations. As an alternative approach, the Imprecise Sample Size Dirichlet Model (ISSDM) is introduced and its properties are studied. The prior distribution over the parameters of a multinomial distribution is the basis to learn Bayesian networks using Bayesian scores. Here, we will show that the ISSDM can be used to learn imprecise Bayesian networks, also called credal networks when all the distributions share a common graphical structure. Some experiments are reported on the use of the ISSDM to learn the structure of a graphical model and to build supervised classifiers.     

Actions in the Category of Precrossed Modules in Lie Algebras

For any object L in the category of precrossed modules in Lie algebras PXLie, we construct the object Act(L), which we call the actor of this object. From this construction, we derive the notions of action, center, semidirect product, derivation, commutator, and abelian precrossed module in PXLie. We show that the notion of action is equivalent to the one given in semi-abelian categories, and Act(L) is the split extension classifier for L. In the case of a crossed module in Lie algebras we show how to recover its actor in the category of crossed modules from its actor in the category of precrossed modules.     

Symmetric semisimple modules of group algebras over finite fields and self-dual permutation codes

A module of a finite group over a finite field with a symmetric non-degenerate bilinear form which is invariant by the group action is called a symmetric module. In this paper, a characterization of indecomposable orthogonal decompositions of symmetric semisimple modules and a criterion for the hyperbolic symmetric modules are obtained, and some applications to the self-dual permutation codes are shown.     

Length two extensions of modules for the Witt algebra

In this paper, we establish an explicit classification of length two extensions of tensor modules for the Witt algebra using the cohomology of the Witt algebra with coe?cients in the module of the space of homomorphisms between the two modules of interest. To do this we extended our module to a module that has a compatible action of the commutative algebra of Laurent polynomials in one variable. In this setting, we are be able to directly compute all possible 1-cocycles.     

Forecasting both time varying MTBF of fighter aircraft module and expected demand of minor parts

A fundamental aspect of inventory management of fighter aircraft is to forecast the demand of minor parts installed in the main module. Most of the existing inventory models do not take into account the time-varying characteristics of modules composed of minor parts even though the demand rate of minor parts is sensitive to the conditions of module. In this paper, we propose random-effects regression models to forecast the mean time between failure (MTBF) of modules and the expected demand for the minor parts in the module based on time-varying characteristics of modules. Subsequently, we show how they can be applied to find necessary seasonal demand of minor parts. Our study is expected to contribute to ROKAF (Republic of Korea Air Force) in terms of providing base information for inventory policy for minor parts.     

Action learning versus strategy learning

This article seeks to ascertain whether the strategy‐learning model of Hanaki, Sethi, Erev, and Peterhansl (2003) better accounts for observed behavior than do the various action‐learning models. It does so by measuring the goodness‐of‐fit of the models' predictions against published experimental results for such games as Coordination, Prisoner's Dilemma, and Chicken. The fit is measured via the mean squared deviation (MSD) between the observed behavior and the one predicted by the model. The results show that, for Chicken, the strategy‐learning model fits the observed data much better than do the action‐learning models. The best action‐learning model, on the other hand, fits the observed data well in Coordination. Overall, the strength of the strategy‐learning model is best shown in games where alternations between the two stage‐game Nash equilibria are often observed in the laboratory experiments. © 2004 Wiley Periodicals, Inc. Complexity 9: 41–50, 2004     

Dynamic modularity in discrete-time models of regulatory networks

We study the properties of small regulatory networks treated as non-autonomous dynamical systems, otherwise called modules when working inside larger networks. We explicit and study the conditions on the input sequences and the internal parameters of the system to behave as a transducer (finite-state automata with inputs and outputs). In the allowed families of networks, we distinguish those with and without feedback on the basis of whether the internal dynamics of the module has a role on determining their input–output behaviors or not. The input–output and non-autonomous bifurcation analysis of this class of modules rely on studying their symbolic dynamics. We consider the interplay between the internal and structural properties of the modules and the different possible inputs on them to deduce possible new functionalities as internal and external responses. Far from the over-optimistic view according to which to a module shall correspond one functionality, we obtain a trade-off between a large spectrum of behaviors and the robustness of each of them depending on the delays, non-linearities and strengths involved in the regulations.     

The graded modules for the graded contact cartan algebras

In this paper we investigate the graded modules for the graded contact Cartan algebras K(n, m) and K(n). For a canonical basis of uPTG module, we derive a commutator formula and then realize Shen's mixed product module in uPTG module ν(n, m) for H(n, m). Considering the Poisson subalgebra K as 1-dimensional central extension of H(n,m), we describe the irreducible PTG modules for K(n,m) and K(n) respectively. In particular, for arbitrary K(n,m), we recover Holmes' work for K(n,1)     

The development of modular-based design in considering technology complexity

Applying modularity in the designing of products has been extensively researched recently to reduce the delay of product development. This paper presents a methodology of modular-based design in the conceptual stage of systems to support concurrent engineering (CE). First, the functions (Fs) are classified into different types of modules according to the correlation in design by using fuzzy cluster identification. Second, the optimal module type is selected based on the considerations of the manufacture and assembly complexities of the system for progressive parallel design. Third, the design priority of Fs within a module is scheduled by measuring the information content of Fs. As a result, the traditional design process is arranged as a series-parallel action to reduce the design time of products. Finally, an automated guided vehicle (AGV) system is used as an example to describe this method.     

Using genetic programming to evolve action selection rules in traversal-based automated software testing: results obtained with the TESTAR tool

Traversal-based automated software testing involves testing an application via its graphical user interface (GUI) and thereby taking the user’s point of view and executing actions in a human-like manner. These actions are decided on the fly, as the software under test (SUT) is being run, as opposed to being set up in the form of a sequence prior to the testing, a sequence that is then used to exercise the SUT. In practice, random choice is commonly used to decide which action to execute at each state (a procedure commonly referred to as monkey testing), but a number of alternative mechanisms have also been proposed in the literature. Here we propose using genetic programming (GP) to evolve such an action selection strategy, defined as a list of IF-THEN rules. Genetic programming has proved to be suited for evolving all sorts of programs, and rules in particular, provided adequate primitives (functions and terminals) are defined. These primitives must aim to extract the most relevant information from the SUT and the dynamics of the testing process. We introduce a number of such primitives suited to the problem at hand and evaluate their usefulness based on various metrics. We carry out experiments and compare the results with those obtained by random selection and also by Q-learning, a reinforcement learning technique. Three applications are used as Software Under Test (SUT) in the experiments. The analysis shows the potential of GP to evolve action selection strategies.     

On the Local Homology Theory for Artinian Modules

Let A be a non-zero Artinian R-module. For an arbitrary ideal I of R, we show that the local homology module Hpx(A) is independent of the choice of x whenever 0 : AI = 0 : A(x1,...,xr). Thus, identifying these modules, we write HpI(A). In this paper we prove that there is a certain duality between HiI(A) and the local cohomology modules and provide some information about the vanishing local homology module HiI(A) which gives an improved form of the main results of [22].     

Module structures and the derived functors of iterated loop functors on unstable modules over the Steenrod algebra

The calculation of the iterated loop functors and their left derived functors on the category of unstable modules over the Steenrod algebra is a non-trivial problem; Singer constructed an explicit and functorial chain complex to calculate these functors. The results of Singer are analysed to give information on the behaviour of these functors with respect to the nilpotent filtration of the category of unstable modules.We show that, if an unstable module M supports an action of an unstable algebra K, then the derived functors of the iterated loop functors applied to M support actions of iterated doubles of K. This allows the finiteness results of Henn on unstable modules which support actions of unstable algebras to be applied to deduce structural results on the derived functors of iterated loops on such modules.     

Discrete neural dynamic programming in wheeled mobile robot control

In this paper we propose a discrete algorithm for a tracking control of a two-wheeled mobile robot (WMR), using an advanced Adaptive Critic Design (ACD). We used Dual-Heuristic Programming (DHP) algorithm, that consists of two parametric structures implemented as Neural Networks (NNs): an actor and a critic, both realized in a form of Random Vector Functional Link (RVFL) NNs. In the proposed algorithm the control system consists of the DHP adaptive critic, a PD controller and a supervisory term, derived from the Lyapunov stability theorem. The supervisory term guaranties a stable realization of a tracking movement in a learning phase of the adaptive critic structure and robustness in face of disturbances. The discrete tracking control algorithm works online, uses the WMR model for a state prediction and does not require a preliminary learning. Verification has been conducted to illustrate the performance of the proposed control algorithm, by a series of experiments on the WMR Pioneer 2-DX.     

Optimal component selection based on cohesion & coupling for component based software system under build-or-buy scheme

Component based software system approach is concerned with the system development by integrating components. The component based software construction primarily focuses on the view that software systems can be built up in modular fashion. The modular design is a logical collection of several independent developed components that are assembled with well defined software architecture. These components can be developed in-house or can be obtained commercially from outside market making build versus buy decision an important consideration in development process. Cohesion and coupling (C&C) plays a major role in determining the system quality in terms of reliability, maintainability and availability. Cohesion is defined as the internal interaction of components within the module. On the other hand, coupling is the external interaction of the module with other modules i.e. interaction of components amongst the modules of the software system. High cohesion and low coupling is one of the important criteria for good software design. Intra-modular coupling density (ICD) is a measure that describes the relationship between cohesion and coupling of modules in a modular software system and its value lies between zero and one. This paper deals with the selection of right mix of components for a modular software system using build-or-buy strategy. In this paper, fuzzy bi-criteria optimization model is formulated for component selection under build-or-buy scheme. The model simultaneously maximizes intra-modular coupling density (ICD) and functionality within the limitation of budget, reliability and delivery time. The model is further extended by incorporating the issue of compatibility amongst the components of the modules. A case study is devised to explain the formulated model.