Interview with Prof. Vincenzo Loia of Universita degli Studi di Salerno, chairman of the 2017 IEEE Symposium on Computational Intelligence in Intelligent Agents (Interviewed by Dr. David Fogel, co-chair, 2017 IEEE Symposium Series on Computational Intelligence.)
Prof. Vincenzo Loia has been actively involved in the computational intelligence community for many years, with numerous publications in the use of fuzzy logic for control, image processing, and clustering. Now, Prof. Loia serves as chairman of the IEEE 2017 Symposium on Computational Intelligence in Intelligent Agents, to be held as part of the 2017 IEEE Symposium Series on Computational Intelligence, held in Honolulu, HI, Nov. 27-Dec. 1. I recently reviewed several pertinent issues of intelligent agent-based modeling with Prof. Loia.
DF: Let’s start by asking “What is an intelligent agent?”
VL: An intelligent agent may be considered an intelligent program that does not suffer from the tyranny of a dictator. By that, I mean the dictatorial instructions of the original programmer. Usually, in software development, the software engineer thinks of everything ahead of time and writes programs to handle every case he or she thinks of. That’s not really the case with intelligent agents. An intelligent agent is typically a software algorithm that is given the duty of proactively trying to meet desired objectives, and it can do this autonomously without human or other machine intervention. Agents work via cooperative strategies, so a single agent does not need to be exhaustive in terms of intelligence or knowledge. The intelligence involved is more of a collective property that is acquired by many agents working together, although each individual agent can also be a learning agent.
DF: What are the main applications of computationally intelligent agents?
VL: The main application domain right now is web intelligence, that is, web-based application is the domain where agents may be put to best use. In particular, if intelligent agents are coupled with semantic web resources (Web 3.0), a wide range of application domains becomes possible even if detached from the Web. Let’s think about the Web for a moment. The Web evolved from connecting information, to connecting people, and just recently to connecting knowledge. This last wave has become possible thanks to intelligent agents that operate, as delegated by humans, on a Web where data are not owned but instead shared, where services show different views for the same web or the data. This is possible when Web resources are handled with declarative ontological languages (i.e.OWL) to produce domain-specific ontologies that agents can use to reason about information and infer new knowledge.
DF: I’ve seen that agent-based modeling can be very useful in better understanding complex adaptive systems. Perhaps you have some of your own favorite examples?
VL: The thing about modeling complexity is that it’s expensive in terms of model representation. In addition, there are often unintended consequences of interacting models. Agent-based technology may be very useful here, because instead of representing a complex system in one complete model, a system of agents can modify their individual behaviors and better reflect the dynamics of the complex system we are interested in. In this way, adaptive dynamic systems are a natural setting (for example, ecosystems) to use intelligent agents in model-based methods.
DF: We hear a lot about deep learning in the news. Does deep learning have a place in agent-based CI?
VL: This is an emergent area of interest within intelligent agents. There’s now some early work in which multi-agent systems use deep reinforcement learning to assess how effectively agent policies can be learned in an environment. In general, deep learning can be very useful in better supporting systems with huge numbers of agents. When agents have high dimensionality, it may be impossible to think about programming them directly. Instead, setting up a learning system – including a deep learning system – can be a successful approach to creating adaptive agents in these problems. One example I should point to is the fruitful synergy between agents and deep learning that’s been reported in Google DeepMind for generative models of images and videos.
DF: We touched on using intelligent agents in ecosystems. What about applying computationally intelligent agents to better understand biological systems?
VL: Agents have been exhaustively used for many artificial life systems – where people set use computational intelligence tools to try to better under the properties of life and living systems. Agent-based modeling is a natural way to represent biological entities. Recently agents are being used not only to represent computational molecular biology but also for computational systems biology.
DF: What are you working on in this area?
VL: My personal and more recent work in the intelligent agents focuses on using agents inside situational awareness systems. In this last year, I’ve given six keynotes on this topic at international conferences.
DF: What led you to this work?
VL: Industrial projects based on situational awareness and requests from a national government agency. I’ve applied agent technology in several industrial projects, most of them funded by the Italian minister of Industry and in collaboration with companies. I started in the 2000’s and worked on agent-based architectures for network inventory, agent-based systems for medical diagnosis, multi-agent architectures for multiple public utilities, semantic agents for automatic document classification, and many others. Since 2012 I’ve been interested in situational awareness (SAW), I’ve combined SAW with agent technology and exploited these results to design decision support systems for smart infrastructure management (for example, see http://mar-te.com/). This project, supported by the Italian Ministry of University and Research, aims at reconfiguring the sea-land logistic processes (especially in Southern Italy) and will define practical solutions to the most critical management and organizational processes of port and freight terminal logistic activities, even in order to reduce the environmental footprint of the sea-land logistics. The scientific results have been presented at IEEE Cogsima in 2015, 2016, and in the forthcoming 2017 conference, and in different keynotes I’ve delivered in the last six months.
DF: My own experience with intelligent agents has been in using evolutionary algorithms to serve as a learning mechanism for competing or cooperating agents in various games, such as the iterated prisoner’s dilemma or the El Farol Game, known more widely as a version of the minority game, or in some cases defense applications.
VL: Yes, it’s interesting to have learning capabilities in diverse applications, and at various levels, within an agent-based system. Recently I’ve been using agents in situational awareness to detect crime by analyzing Web resources.
DF: What do you think someone not working directly in intelligent agents would gain by attending your symposium?
VL: Agents are my view related to many facets of computational intelligence so they should attract non-specialists who are interested in application domains rather than agent technology itself. I think people with problems that involve complex adaptive systems should be very interested in the agent-based technologies that will be described at the symposium.
Vincenzo Loia: University of Salerno, Dept. Management and innovation Systems, Salerno Italy, email@example.com
The URL for the 2017 IEEE Symposium Series on Computational Intelligence is forthcoming.
To read more:
Giuseppe D’Aniello, Vincenzo Loia, Francesco Orciuoli, “A multi-agent fuzzy consensus model in a Situation Awareness framework” Applied Soft Computing, 30: 430-440, 2015.
Giuseppe D’Aniello, Angelo Gaeta, Vincenzo Loia, Francesco Orciuoli, “Integrating GSO and SAW ontologies to enable Situation Awareness in Green Fleet Management”, 2016 IEEE International Multi-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), pp. 138-144, 2016, ISSN 2379-1675.
(C) David Fogel, 2017