Gaia - A Software Framework for Temporal Probabilistic Reasoning
P. Wiggers, L.J.M. Rothkrantz
Key Words:
Temporal probabilistic reasoning; (dynamic) Bayesian networks; speech and language processing.
Abstract:
In this paper we present our general purpose framework for probabilistic temporal reasoning that is specifically designed for efficient processing in domains with large state spaces. We discuss the design and the features of the toolkit. Features include: sparse representations of probability tables, a fast algorithm for inference with probability tables, lazy evaluation of probability tables, algorithms for calculations with treeshaped distributions, the ability to change distributions on the fly, and a generalisation of dynamic Bayesian networks that we developed. finally, we discuss the construction of a speech recognizer with the toolkit that was used to validate our implementation.

e-Learning of Phase Change Processes under Vigorous Convection Heat Transfer
S. Kimura, K. Kanev
Key Words:
Heat transfer; phase change; solidification; convection; e-learning; GIFES.
Abstract:
In this work we discuss dynamic behaviors of solidification processes and formulate a one-dimensional model for their analysis. Based on it a fast and efficient code for numerical simulations and solidification layer thickness predictions has been implemented. A specialized experimental apparatus for observation and measurement of time varying ice layer thickness under different cooling conditions has been constructed and employed for gathering of experimental data. An extensive analysis and comparisons of model-based nimerical results and experimentally measured layer thickness have been conducted. Following the very good agreement of simulated and measured results we are integrating the numerical simulation of solid layer dynamic responses to cooling temperature modulations and convective fluid flow deviations into traditional engineering courses. We are also focusing on e-learning where access to experimental facilities is much more limited and modelling and numerical simulations can play a central role. In the context of e-learning we discuss the registration and employment of the developed numerical code into the specialized highly configurable Graphical Interface Framework for Educational Support (GIFES).

Design of Advanced Computer Architectures, Based on PIM - Processors in Memory
T.Tashev, S.Tashev, N. Tasheva
Key Words:
Processor in memory; cache coherent protocols; multiprocessing; performance evaluation.
Abstract:
The present article describes the design and simulation of computer architecture based on PIM (Processor-In-Memory). PIM models are analyzed according to the developed mathematical formulations used for simulation of the proposed architecture as a possibility to solve the bottleneck problem between the main processor and memory, providing high bandwidth and computational resources in the operating memory. The proposed architecture was analysed for the performance measuring and time markers researches with Processor-In-Memory (PIM) computing.

A Polynomial Algorithm for Interval Minimum Spanning Tree Problem
A. Hossain
Key Words:
Interval algorithm; interval minimum spanning tree problem; network; uncertainty.
Abstract:
A new polynomial algorithm is presented for solving the minimum spanning tree problem. The algorithm is applicable to the case when the generalized distance associated with each arc is non-negative, interval or real. The interval algorithm is developed on the base of midpoint and half-width representation of intervals and is more efficient than the interval algorithm that could be proposed by using traditional interval description. The complexity of the new algorithm is evaluated.

Adaptive Stabilizing Controller Design Performing under Stability of Sliding Type
S.Stanchev
Key Words:
Design method; Direct adaptive stabilizer; Sliding stability; Robustness; Inverted pendulum benchmark.
Abstract:
This work studies an application of redesign of classical MRAC controller for nonlinear plant allowing description in canonical nonlinear form. Since MRAC scheme employs a linearized model of the plant in a state space form, focused not only on robustness, but first on stability analysis of the system. The siplicity of this redesign having in mind repeating of the main steps of considered classical design with all advantages of the method is compensated by introducing another stability type which is more reliable and mathematically simpler. The manner of using of sliding stability removes its main disadvantage - the chattering phenomenon. Chattering disappears because in this redesign the use of variable structure is avoided, since, the Lyapuov derivative of classical design is modified for guaranteeing sliding stability.