Selected Publications
(this part requires updating!)
ARAUJO, Ricardo Matsumura de; LAMB, Luis
C. Memetic Networks: Analyzing the Effects of Network
Properties in Multi-Agent
Performance. In:
Proceedings of the Twenty-Third AAAI Conference on Artificial
Intelligence (AAAI-08), 2008, Chicago, USA.
We explore the
relationship between properties of the network defined by connected
agents and the global system performance. This is achieved by means of
a novel class of optimization algorithms. This new class makes explicit
use of an underlying network that structures the information flow
between multiple agents performing local searches. We show that this
new class of algorithms is competitive with respect to other
population-based optimization techniques. Finally, we demonstrate by
numerical simulations that changes in the way the network is built
leads to variations in the system's performance. In particular, we show
how constrained hubs - highly connected agents - can be beneficial in
particular optimization problems.
ARAUJO, Ricardo Matsumura de; LAMB, Luis
C. An information theoretic analysis of memory bounds in
a distributed resource allocation mechanism. In:
Proceedings of the 20th International Joint Conference on Artificial
Intelligence (IJCAI 2007), 2007, Hyderabad, India.
Multiagent
distributed resource allocation requires that agents act on limited,
localized information with minimum communication overhead in order to
optimize the distribution of available resources. When requirements and
constraints are dynamic, learning agents may be necessary to allow for
adaptation. One way of accomplishing learning is to observe past
outcomes and use this information in order to improve future decisions.
When limits in agents' memory or observation capabilities are assumed,
one must decide on how large should the observation window be. We
investigate how this decision influences both agents' and system's
performance in the context of a special class of distributed resource
allocation problems, namely dispersion games. Our contribution is
twofold. First, we show by numerical experiments over a specific
dispersion game (the Minority Game) that in such scenario an agent's
performance is non-monotonically correlated with her memory size when
all other agents are kept unchanged. Second, we provide an
information-theoretic explanation for the observed behaviors, showing
that a downward causation effect takes place.
ARAUJO, Ricardo Matsumura de; LAMB, Luis C. On
the Evolution of Memory Size in the Minority Game (Extended Abstract).
In: Proceedings of the 19th International Joint Conference on
Artificial Intelligence (IJCAI 2005), 2005, Edinburgh, Scotland.
ARAUJO, Ricardo Matsumura de ; LAMB, Luis C. Sobre
Tamanhos de Memória no Minority Game. In: Encontro
Nacional de Inteligência Artificial (ENIA), 2005, Sao Leopoldo, RS.
PRIMO, Alex Fernando Teixeira ; RECUERO, Raquel da Cunha ;
ARAUJO, Ricardo Matsumura de. The Co-link Project:
collaborative writing of multidirectional links. In: ACM
Conference on Computer Supported Cooperative Work (CSCW), 2004,
Chicago, Illinois.
After
reviewing issues about the politics of links and open hypertext
systems, Co-link project is introduced (
http://www.co-link.org).
Its capability of allowing the collaborative writing of
multidirectional links is described and its possible impact on
education and research is suggested.
ARAUJO, Ricardo Matsumura de; LAMB, Luis C. Neural-Evolutionary
Learning in a Bounded Rationality Scenario. In: 11th
International Conference on Neural Information Processing (ICONIP),
2004, Calcutta, India.
This paper
presents a neural-evolutionary framework for the simulation of market
models in a bounded rationality scenario. Each agent involved in the
scenario make use of a population of neural networks in order to make a
decision, while inductive learning is performed by means of an
evolutionary algorithm. We show that good convergence to the
game-theoretic equilibrium is reached within certain parameters.
ARAUJO, Ricardo Matsumura de; LAMB, Luis C. Towards
Understanding the Role of Learning Models In the Dynamics of the
Minority Game. In: 16th IEEE International Conference
on Tools with Artificial Intelligence (ICTAI), 2004, Boca Raton, FL.
This paper
reports experiments in a boundedly rational evolutionary game, namely
the Minority Game, where agents apply a very simple learning algorithm
to discard bad strategies and create new ones. The results show that
even such simplified learning model presents qualitative differences
from the behaviour of the traditional game, where strategies are fixed
and cannot be modified or discarded. We show that this results are
qualitatively similar to other, more complex, learning approaches.
Also, we study how the learning parameters of our model affects the
dynamics of the game, evidencing a high dependence between the
behaviour of the system and the way fitness is attributed to new
strategies entering the game.
BOTELHO,
Silvia Silva da Costa; ARAÚJO, Ricardo Matsumura de; TADDEI,
Lorenzo; PELLEJERO, David; NEVES, Renato; COSTA, Rodrigo Mendes.Furgbol
- Construindo Robôs Autônomos Holonômicos Para Jogar
Futebol. In: Simpósio Brasilerio de Automação
Inteligente (SBAI), 2003, Bauru, SP.
This paper
presents a prototype of an autonomous mobile robot to play soccer and
to test artificial intelligence theories. We propose a architecture
that allows an autonomous behavior, avoiding obstacles and reaching the
goals. Topics associated with the prototype implementation are
presented. We analyze the result of a set of tests, validating our
system.
Aprendizado de máquina em sistemas complexos
multiagentes: estudo de caso em um ambiente sob racionalidade limitada.
Thesis for the master's degree in Computer Science, at UFRGS. 2004.
This work
investigates the relationship between learning and dynamics in complex
multiagent systems. We do so by means of experimental studies in a
bounded rationality scenario which lays at the intersection of
Artificial Intelligence, Economics and Statistical Physics known as the
"Minority Game". We present experimental results about the game aiming
at studying such scenario under a Machine Learning perspective. We
introduce a new learning algorithm for agents in the Minority Game,
namely creative learning, and show that the algorithm renders a more
efficient distribution of resources among agents. This increase in
efficiency is shown to be a result of an unrestricted search in the
strategies space which allows for an efficient maximisation of the
distance between agents strategies. We then analyse the effects of this
algorithm's parameters in the performance of an agent, comparing the
results with the traditional learning algorithm, concluding that the
proposed algorithm is more efficient than the traditional one in most
situations. Finally, we investigate how memory size affects agent's
performance using both algorithms, showing that individual agents with
larger memory sizes achieve better performances only when the system as
a whole is at an inefficient phase, while in other phases such
increases are irrelevant - and even harmful - to the performance of the
agents.
Projeto e implementação de um robô holonômico
para futebol de robôs. Dissertation presented to obtain
the Computer Engineering degree at FURG. 2002.
