
Fernando FernándezUniversity Carlos III de Madrid | UC3M · Department of Computer Science and Engineering
Fernando Fernández
Ph.D. Computer Science
About
132
Publications
39,921
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
2,228
Citations
Introduction
Fernando Fernández currently works at the Department of Computer Science and Engineering, University Carlos III de Madrid. Fernando does research in Artificial Intelligence, specifically in Machine Learning and Automated Planning, and their application to real systems like social robotics
Additional affiliations
October 2004 - December 2005
Publications
Publications (132)
SkeletonAgent is an agent framework whose main feature is to integrate different artificial intelligent skills, like planning or learning, to obtain new behaviours in a multi-agent environment. This framework has been previously instantiated in a deliberative domain (electronic tourism), where planning was used to integrate Web information in a tou...
Policy Reuse is a method to improve reinforcement learning with the ability to solve multiple tasks by build- ing upon past problem solving experience, as accumu- lated in a Policy Library. Given a new task, a Policy Reuse learner uses the past policies in the library as a probabilistic bias in its new learning process. We present how the effective...
Instance Based Methods for classification are based on stor-ing the complete training dataset. Once a query is received, it is compared with all the instances in the dataset, providing an answer as a function of the labels of the most similar instances. Opposite to this, Nearest Prototype Classification (NPC) obtains in training time a reduced set...
In the probabilistic track of the last International Plan-ning Competition two main approaches were used, Markov Decision Processes (Boutilier, Dean, & Hanks 1998) and decision-theoretic planning (Blythe 1999). Both approaches use a domain representation with an explicit definition of the probabilities of the actions ef-fects. But when planning in...
Learning a pedagogical policy in an Adaptive Educational System (AIES) fits as a Reinforcement Learning (RL) problem. However, to learn pedagogical policies requires to acquire a huge amount of ex- perience interacting with the students, so applying RL to the AIES from scratch is infeasible. In this paper we describe RLATES, an AIES that uses RL to...
Classical planning domain representations assume all the ob- jects from one type are exactly the same. But when solving problems in the real world systems, the execution of a plan that theoretically solves a problem, can fail because of not properly capturing the special fea- tures of an object in the initial representation. We propose to capture t...
When executing plans in real world, a plan that theoretically solves a problem, can fail because of special features of an object were not properly captured in the initial domain representation. We propose to capture this uncertainty about the world repeating cycles of planning, execution and learning. In this paper, we describe the Planning, Execu...
Proceeding of: Algorithmic learning theory, 15th International Conference on Algorithmic Learning Theory (ALT 04), Padova, Italy, October 2-5, 2004 One of the most important issues in educational systems is to define effective teaching policies according to the students learning characteristics. This paper proposes to use the Reinforcement Learning...
In pattern classification problems, many works have been carried out with the aim of designing good classifiers from different perspectives. These works achieve very good results in many domains. However, in general they are very dependent on some crucial parameters involved in the design. These parameters have to be found by a trial and error proc...
10 pages, 4 figures.-- Contributed to: Workshop on Human Computer Interface for Semantic Web and Web Applications (HCI-SWWA’03). Part of the International Federated Conferences (OTM'03, Catania, Sicily, Italy, Nov 3-7, 2003). The paper shows the architecture of the RLATES system, an Adaptive and Intelligent Educational System that uses the Reinforc...
The definition of effective peda gogical strategies for coaching and tutoring students ac- cording to their needs is one of the most important issues in Adaptive and Intelligent Educational Systems (AIES). The use of a Reinforcement Learning (RL) model allows the system to learn au- tomatically how to teach to each student individually, only based...
The design of nearest neighbour classifiers is very dependent from some crucial parameters involved in learning, like the
number of prototypes to use, the initial localization of these prototypes, and a smoothing parameter. These parameters have
to be found by a trial and error process or by some automatic methods. In this work, an evolutionary app...
Neural networks have proven to be very powerful techniques for solving a wide range of tasks. However, the learned concepts are unreadable for humans. Some works try to obtain symbolic models from the networks, once these networks have been trained, allowing to understand the model by means of decision trees or rules that are closer to human unders...
When applying a Reinforcement Learning technique to problems with continuous or very large state spaces, some kind of generalization is required. In the bibliography, two main approaches can be found. On one hand, the generalization problem can be de- fined as an approximation problem of the continuous value function, typically solved with neural n...
In this paper we propose the use of a Reinforcement Learning model that avoids the problems derived from construction of different pedagogic strategies for each student in intlligent tutoring systems. This model consists on defining the pedagogical module of any ITS (domain independent) as an action policy learned by trial and error (as human tutor...
The design of nearest neighbour classifiers can be seen as the partitioning of the whole domain in different regions that can be directly mapped to a class. The definition of the limits of these regions is the goal of any nearest neighbour based algorithm. These limits can be described by the location and class of a reduced set of prototypes and th...
Reinforcement learning has proven to be a set of successful techniques for finding optimal policies on uncertain and/or dynamic
domains, such as the RoboCup. One of the problems on using such techniques appears with large state and action spaces, as
it is the case of input information coming from the Robosoccer simulator. In this paper, we describe...
Policy Reuse is a reinforcement learning method in which learned policies are saved and reused in similar tasks. The policy reuse learner extends its exploration to probabilis-tically include the exploitation of past poli-cies, with the outcome of significantly im-proving its learning efficiency. In this paper we demonstrate that Policy Reuse can b...
REPLICA is a relational instance based learning module for solving STRIPS planning problems described in PDDL. REPLICA learns a reduced policy represented by a set of pairs <meta-state,action>. The meta-state represents the current planning state and the goal; the action represents the opera-tor to execute in such meta-state. Both are described in...
Data mining is a difficult task that relies on an exploratory and analytic process of large quantities of data in order to discover meaningful patterns and rules. It requires complex methodologies, and the increasing heterogeneity and com- plexity of available data requires some skills to build the data mining processes, or knowledge flows. The goa...
Resumen El principal problema a que nos enfrentamos al diseñar sistemas multi-agente es cómo coordinar los agentes que pertenecen a este sistema para obtener un comportamiento global eficiente. En algunos trabajos, el comportamiento coordinado de los agentes se obtiene gracias a un conocimiento total del dominio en el que se pueden aplicar planific...
Resumen La coordinación emergente pretende obtener comportamientos colaborativos entre diversos agentes sin que eso implique que cada individuo deba tener un conocimiento global del dominio, y sin que ese conocimiento deba estar centralizado. Al no requerir conocimiento global, se minimiza la comunicación entre los agentes de forma que cada uno de...
El aprendizaje por refuerzo es un modelo de aprendizaje que permite implementar comportamientos inteligentes de forma automática. La mayor parte de la teoría del aprendizaje por refuerzo tiene su fundamento en la programación dinámica, y por tanto, en lo que se denominan funciones de valor. Sin embargo, la implementación tradicional de estas funcio...