Catálogo de publicaciones - libros

In many applications software agents are supposed to show adaptive behaviour and learning capabilities in information rich environments. On the other hand agents are often expected to be resource-bounded systems, which do not utilize much memory, disk space and CPU time. In this paper we present a novel framework for incremental, statistical learning, attempting to satisfy both requirements. The new method, called APS, runs in a cycle including such phases as: storing observations in a history, rule discovery using data mining algorithms, and knowledge base maintenance. Once processed, the old facts are removed from the history and in every subsequent learning run only the recent portion of observations is analysed in search of new rules. This approach can substantially save disk space and processing time as compared to batch learning methods.

It is widely believed, that the expressivity of STRIPS and STRIPS-like planning based on actions is generally lower than the expressivity of Hierarchical Task Network (HTN) and HTN-like planning, based on hierarchical decomposition. This would mean that a HTN-like planner can generally solve more domains than a STRIPS-like planner with the same extensions. In this paper, we show that both approaches, as they are practically used, are identically expressive and can solve all domains solvable by a Turing machine with finite tape (i.e. solvable by a common computer).

The paper contains a description of the implementation and performance evaluation of the agent-based population learning algorithm used to train the feed-forward artificial neural networks. The goal of the research was to evaluate efficiency of the agent-based approach and to establish experimentally which different factors representing the A-team structure and topology affect the performance of the analyzed agent-based algorithm. The paper includes a general overview of the JABAT environment used to deploy the ANN training algorithm, a description of different agents employed and their roles, as well as the computational experiment plan and the discussion of the performance evaluation results.

Service discovery in ubiquitous environments faces various issues such as dynamic-topology, devices-capability, resource-sharing, interoperability etc. In this paper, we propose Service Discovery Middleware (SDM) to solve the interoperability issue by using Discovery Middleware (DM), which was introduced by the FIPA (Foundation for Intelligent Physical Agents) [1]. The SDM provides an appropriate choice for providing interoperability in heterogeneous environments. Moreover, it supports a means for autonomic discovery of heterogeneous integration services in ubiquitous computing and transparency between users and services. With these ideas in mind, we design a simple mechanism for building a wide range of distributed services and applications as well as for supporting a softness and adaptable means to control and manage the Service Discovery Protocols (SDP). Furthermore, we implement as an example, a Personal Agent (PA) platform based on the FIPA-OS agent platform, in order to provide SDM component functionality [10] [20].

In this paper, the performance between RFC compatible normal TCP and several speed constraints ignored fast TCP is compared. To do these, the main algorithms that constraints the transmit rate of TCP are removed. We, and also, have modified TCP protocol stack in a Linux kernel as a kind of agent system to compare the speeds between the standard TCP and our modified fast TCP. We find that if the destination agent is short distance away from the source agent and packet error is scarce then the speed differences between normal and fast TCP may be negligible. However, if the destination agent is far away from the source agent and slow start algorithm is not adopted then the transfer time for small file is different greatly. In addition, if packet error occurred frequently, our modified fast TCP is faster than the standard TCP regardless of distance.

In this paper we introduce first order Paraconsistent Annotated Multimodal systems M which may constitute a framework for multi-agent system reasoning. Such systems are capable of handling imprecise, inconsistent and paracomplete knowledge in a non-trivial manner in their structures.

Multi-agent systems have been attacking the challenges of distributed information retrieval. In this paper, we propose a consensus method-based framework to evaluate the performance of cooperative information retrieval tasks of the agents. Two well-known measurements, and , are extended to handle consensual closeness (i.e., local and global consensus) between the retrieved results. We show in a motivating example that the proposed criteria are prone to solve the problem of rigidity of classical and . More importantly, the retrieved results can be ranked with respect to the consensual score.

This paper proposes a hierarchy of logics of agents relative to levels of their conflicts, self-conflicts and irrationality to provide a base for several studies on foundations of the theories of uncertainties. These studies include the foundation of known uncertainty theories (probability theory, fuzzy logic, and others) as well as new logics and types of uncertainties. Probability theory, fuzzy logic, and others theories of uncertainties provide a calculus for manipulating with probabilities, membership functions, and other types of uncertainty indicators. However, these theories lack a mechanism for getting initial (basic) uncertainties. The proposed hierarchy of conflicting and irrational agents creates a base for generating uncertainty values, logic operations with these values and for comparing different types of uncertainty for the preference relation. A core concept of this hierarchy is the concept of expansion by superposition that includes fusion and adjustment of contradictory events and statements.

In stratified multi-agent planning (SMAP), the parent planning agent and its child planning agents work together to achieve a goal. The parent planning agent executes a rough plan for a goal, and the child planning agents execute detailed plans for subgoals. Although this kind of SMAP is efficient, it is difficult for the parent agent to change the plan while a child planning agent is working. On the other hand, on-line planning, where the agent continuously updates its plan during the plan execution, is very important if we need to implement planning agents working in dynamic environments. This paper shows how to realize on-line planning in SMAP. For this purpose, we extend Dynagent which is an on-line HTN planning agent system.

Web services selection is based on QoS and trust. As one of the important attributes of QoS, reputation is commonly used to assess the trustworthiness of the web services and minimize the threats of transactions. However, most existing reputation models of web services are all based on the subjective user ratings. These systems are easily attacked by malicious raters. This paper presents a novel reputation model named WSrep, in WSrep, the reputation integrates user ratings and a significant objective factor-credibility of QoS advertisements which is an objective view of the past behaviors of a given service. Other contributions of the paper include a customer measurable QoS model, a Bayesian learning model for building the credibility, and a set of experiments to show the benefits of our approach.