Advanced computational models and learning theories for spoken language processing

Recent developments in research on humanoid robots and interactive agents have highlighted the importance of and expectation on automatic speech recognition (ASR) as a means of endowing such an agent with the ability to communicate via speech. This article describes some of the approaches pursued at NTT Communication Science Laboratories (NTT-CSL) for dealing with such challenges in ASR. In particular, we focus on methods for fast search through finite-state machines, Bayesian solutions for modeling and classification of speech, and a discriminative training approach for minimizing errors in large vocabulary continuous speech recognition.     

Complementing computational fluid dynamics methods with classical analytical techniques

New aerospace vehicle designs must have greater performance and versatility at affordable cost. This requires multi-disciplinary analysis and optimization which in turn requires more accurate and efficient numerical simulation tools. The need for greater accuracy and efficiency of computational fluid dynamics (CFD) tools is further amplified by the industry trend toward distributed computing (e.g. workstation clusters) and away from supercomputers. Complementary analytic methods coupled with traditional CFD approaches offer the means for increased simulation capability by incorporating more essential physics into solution algorithms and reducing reliance on grid density for achieving accuracy. McDonnell Douglas Aerospace has a focused activity directed at improving affordability of CFD tools with complementary analytic techniques and has developed a strong capability. Results have proven very successful. Several examples of ongoing work are discussed, including improved far-field boundary conditions for CFD codes and analytic-based aerodynamic analysis and design optimization methods.     

Combining seasonal ARIMA models with computational intelligence techniques for time series forecasting

Seasonal autoregressive integrated moving average (SARIMA) models form one of the most popular and widely used seasonal time series models over the past three decades. However, in several researches it has been argued that they have two basic limitations that detract from their popularity for seasonal time series forecasting tasks. SARIMA models assume that future values of a time series have a linear relationship with current and past values as well as with white noise; therefore, approximations by SARIMA models may not be adequate for complex nonlinear problems. In addition, SARIMA models require a large amount of historical data to produce desired results. However, in real situations, due to uncertainty resulting from the integral environment and rapid development of new technology, future situations must be forecasted using small data sets over a short span of time. Using hybrid models or combining several models has become a common practice to overcome the limitations of single models and improve forecasting accuracy. In this paper, a new hybrid model, which combines the seasonal autoregressive integrated moving average (SARIMA) and computational intelligence techniques such as artificial neural networks and fuzzy models for seasonal time series forecasting is proposed. In the proposed model, these two techniques are applied to simultaneously overcome the linear and data limitations of SARIMA models and yield more accurate results. Empirical results of forecasting two well-known seasonal time series data sets indicate that the proposed model exhibits effectively improved forecasting accuracy, so that it can be used as an appropriate seasonal time series model.     

Customized computational robot dynamics

In 1983 the authors implemented the computer program Algebraic Robot Modeler (ARM) to generate symbolically complete closed-form and recursive dynamic robot models.^1–3 Then, in 1985, we incorporated in ARM heuristic rules for the systematic organization of dynamic robot models to reduce the computational requirements of customized forward and inverse dynamics calculations. We resolve the issue of numerical efficiency of customized closed-form and recursive forward and inverse dynamics algorithms for kinematically and dynamically structured manipulators. We find that ARM-generated customized closed-form algorithms are the most computationally efficient calculators of forward and inverse dynamics of three degree-of-freedom manipulators. For six DOF predominantly rotational manipulators, ARM-generated customized recursive algorithms are the most computationally efficient foward and inverse dynamics algorithms; inverse dynamics can be computed in less than one millisecond on commercially-available processors (in software, without special-purpose hardware). In our companion article,⁴ we compare the symbolic efficiencies of six robot dynamics formulations for generating closed-form and recursive models.     

Generic models for computational effects

A Freyd  -category is a subtle generalisation of the notion of a category with finite products. It is suitable for modelling environments in call-by-value programming languages, such as the computational λ$\lambda$-calculus, with computational effects. We develop the theory of Freyd-categories with that in mind. We first show that any countable Lawvere theory, hence any signature of operations with countable arity subject to equations, directly generates a Freyd-category. We then give canonical, universal embeddings of Freyd-categories into closed Freyd-categories, characterised by being free cocompletions. The combination of the two constructions sends a signature of operations and equations to the Kleisli category for the monad on the category Set   generated by it, thus refining the analysis of computational effects given by monads. That in turn allows a more structural analysis of the _λc${\lambda }_c$-calculus. Our leading examples of signatures arise from side-effects, interactive input/output and exceptions. We extend our analysis to an enriched setting in order to account for recursion and for computational effects and signatures that inherently involve it, such as partiality, nondeterminism and probabilistic nondeterminism.     

Intelligent computational techniques for multimodal data

This guest editorial introduces the special issue on “Intelligent Computational Techniques for Multimodal Data”. The goal of this special issue was to solve a variety of real-life problems which have uncertainty, imprecision, vagueness, resulting in high performance applications or prototypes for real time system. The special issue touched different hot topics related to Computer Vision, Computational Biology, Multimedia data mining, High-dimensional multimedia data, Deep convolution network, Deep semantic preserving hashing, High-dimensional multimedia classification, Deep CNN and extended residual units, particle swarm optimization, Cyberbullying detection on social multimedia, Multimedia detection algorithm of malicious nodes, Multimedia based fast face recognition, Diversifying personalized mobile multimedia application, Multimedia image edge extraction algorithm, Rumour veracity detection.

     

Symbolic computational techniques for solving games

Games are useful in modular specification and analysis of systems where the distinction among the choices controlled by different components (for instance, the system and its environment) is made explicit. In this paper, we formulate and compare various symbolic computational techniques for deciding the existence of winning strategies. The game structure is given implicitly, and the winning condition is either a reachability game of the form p until q (for state predicates p and q) or a safety game of the form Always p.For reachability games, the first technique employs symbolic fixed-point computation using ordered binary decision diagrams (BDDs) [9]. The second technique checks for the existence of strategies that ensure winning within k steps, for a user-specified bound k, by reduction to the satisfiability of quantified boolean formulas. Finally, the bounded case can also be solved by reduction to satisfiability of ordinary boolean formulas, and we discuss two techniques, one based on encoding the strategy tree and one based on encoding a witness subgraph, for reduction to Sat. We also show how some of these techniques can be adopted to solve safety games. We compare the various approaches by evaluating them on two examples for reachability games, and on an interface synthesis example for a fragment of TinyOS [15] for safety games. We use existing tools such as Mocha [4], Mucke [7], Semprop [19], Qube [12], and Berkmin [13] and contrast the results.     

Asymptotic behavior and best approximation in computational fluid dynamics

Computational fluid dynamics has many successes for the solution of simple standard problems. For relatively complex problems, especially if nonlinear and of mixed type, the computed approximate solutions are mostly of dubious accuracy and credibility. The difficulty appears fundamental. Model studies in one space dimension suggest that most of such discrete problems are poorly posed. The sequence of computed solutions at successively refined meshes need not converge; and apparently “smooth” computed approximations can “converge” to wrong limits with large global errors. For certain discrete formulations the sequence is asymptotic in the sense of displaying minimum error at some fairly large critical mesh Reynolds number (coarse meshes). This error minimum can be as small as those promised by the correct “convergent approximations” at much smaller meshes. Certain behavior of the computed solutions around such a critical mesh Reynolds number help to identify the “best approximation”. Such analytic inferences have been tested and verified in the computational solutions of successively more complex flows governed by Navier-Stokes equations in two space dimensions. The flow fields due to shockwave-laminar-boundary layer interaction were computed with different discrete formulations and various perturbations. The computed “best approximations” differ little and all compare favorably with available experimental data. Some of such formulations give the “best approximations” at reasonably coarse meshes, requiring much smaller computational effort; and should therefore be favorably considered.     

Coupling multibody dynamics and computational fluid dynamics on 8192 processor cores

This paper describes a method for the fully resolved simulation of particle laden flows. For this purpose, we discuss the parallelization of large scale coupled fluid structure interaction with up to 37 million geometrically modeled moving objects incorporated in the flow. The simulation is performed using a 3D lattice Boltzmann solver for the fluid flow and a so-called rigid body physics engine for the treatment of the objects. The numerical algorithms and the parallelization are discussed in detail. Furthermore, performance results are presented for test cases on up to 8192 processor cores running on an SGI Altix supercomputer. The approach enables a detailed simulation of large scale particulate flows that are relevant for many industrial applications.     

Advanced interaction techniques in virtual environments

Fundamental to much of the Virtual Reality work is, in addition to high-level 3D graphical and multimedia scenes, research on advanced methods of interaction. The “visitor” of such virtual worlds must be able to act and behave intuitively, as he would in everyday situations, as well as receive expectable natural behaviour presented as feedback from the objects in the environment, in a way that he/she has the feeling of direct interaction with his/her application. In this paper we present several techniques to enrich the naturalness and enhance the user involvement in the virtual environment. We present how the user is enabled to grab objects without using any specific and elaborate hand gesture, which is more intuitive and close to the way humans are used to do. We also introduce a technique that makes it possible for the user to surround objects without any force-feedback interaction device. This technique allows the user to surround or “walk” with the virtual hand on the object's surface and look for the best position to grab it.     

Application of computational intelligence techniques in active networks

 Computational intelligence techniques have been successfully used for solving control problems in packet-switching network architectures. The introduction of active networking adds a high degree of flexibility in customizing the network infrastructure and introducing new functionality. Therefore, there is a clear need for investigating both the applicability of computational intelligence techniques in this new networking environment, as well as the provisions of active networking technology that computational intelligence techniques can exploit for improved operation. We report on the characteristics of these technologies, their synergy and on outline recent efforts in the design of a computational intelligence toolkit and its application to routing on a novel active networking environment.     

Shape optimization with computational fluid dynamics

In many product design and development applications, computational fluid dynamics CFD has become a useful analytical simulation tool. CFD simulations are quite useful in predicting several response parameters for a given design condition. However, like any analysis tool CFD simulations provide limited insight into the design space and the changes needed to find the optimum design parameters.This paper deals with the shape optimization of fluid flows using CFD and numerical optimization techniques. By integrating a commercial optimization code with a CFD code, a CFD shape optimization tool was developed. To study the effectiveness of the developed tool and its ability to produce results with reasonable CPU time, the shape optimization of an airfoil and S-shaped duct are studied with different numbers of design variables. The developed shape optimization tool along with the optimization and CPU time results are discussed.     

Visualization's role in analyzing computational fluid dynamics data

Visualization has fueled the growth and understanding of many scientific and engineering fields. In computational fluid dynamics, for example, engineers now use numerical calculations to accurately simulate many engineering problems that once required the use of physical experiments involving wind and water tunnels. CFD has come to serve as an instrument in the design of many familiar engineering processes. These developments have produced massive amounts of data for analysis. Feature detection has become a critical tool in understanding key structures within these volumes of data. We briefly discuss the growth of CFD and investigate why feature detection and visualization have become such a vital part of analyzing CFD simulation results.