The Artificial Intelligence Lab, founded in 1983 by Prof. Dr. Luc Steels, is the first AI lab on the European mainland. It is headed by Prof. Dr. Ann Nowé and Prof. Dr. Bernard Manderick. During its history of more than three decades, the VUB AI Lab has been following two main routes towards the understanding of intelligence: both the symbolic route (classical AI) as the dynamics route (complex systems science). Academic track record Since its foundation, more than 50 people have received a PhD degree at the Artificial Intelligence Lab, of which 18 in the past 5 years only. The total amount of publications sums up to more than 850 publications which are cited more than 24.000 times, 8500 times in the last 5 years. The lab also has experience in setting up of spin-offs (5 since its beginning). The research group is provided with standard hardware and software utilities and departmental services (secretary and ICT support). The COMO group of the AI-lab focuses on the one hand on the modeling of natural phenomena, and on the other hand on developing algorithms for complex problem solving inspired by these natural phenomena. The lab has experience in a wide range of learning techniques such as: Reinforcement Learning, Genetic Algorithms, Neural Networks, Support Vector Machines, Graphical models including Bayesian Networks, Genetic algorithms etc. Industrial Cooperation Experience The AI lab has applied artificial intelligence techniques including machine learning in a wide range of settings in the past, from robotics (MIRAD – Vlaio SBO), Industry 4.0 (“OperatorInfo”) to epidemics (“Learning optimal preventive strategies to mitigate epidemics of latent infectious diseases” (FWO), in which individual based models are combined with multi-armed bandit problems to come with an ideal disease mitigation strategy. National and international network On the national level the lab collaborates with several research groups. For the proposed project the most relevant ones are: CIB KULeuven, Flanders Make (formerly FMTC), KULeuven Division of Production Engineering, Machine Design & Automation (PMA), INTEC UGent, Systems UGent, PATS UA, Applied Mathematics and Computer science UGent, Distrinet KULeuven, Mebios KULeuven. The lab is also member of the BruBotics Research Center, a VUB initiative which assembles interdisciplinary groups working on all aspects of robotics. The group also has a strong collaboration with their counterparts at the Université Libre de Bruxelles on topics related to Computational Biology and (Evolutionary) Game Theory. The group participates in a new Brussels institute for Bioinformatics research; the new institute called IB² gathers all medical and computational expertise in the Brussels area to tackle medical questions related to cancer, diabetes, etc. Through our alumni we have good connections with industry. They have taken jobs in AI companies like Google and DeepMind, and have been attracted by industry, often through contacts via SBO projects, to set up a team on data analytics (e.g Yann-Michael De Hauwere at Persgroep, Maarten Devillé at Medialaan and Maarten Peeters Forecasting Team Coordinator at ENGIE). There are also many international collaborations including: University of Minho and CFTC Lisbon, University of Maastricht (Kurt Driessens), University of York (Daniel Kudenko), Goldsmiths College London (Mark d’Inverno), Universität Wien (Tecumseh Fitch), University of Haifa (Wendy Sandler), Washington State University (Matt Taylor) and University of Liverpool (Karl Tuyls), and institutes in developing countries, in particular: University of Santa Clara and University of Camaguey in Cuba and University of Moi Kenya.
Research Areas
LEARNING IN MULTI-AGENT SYSTEMS
Telecommunications, economics, mobile robots, traffic simulation, electricity grids/smart grids … are all examples of systems in which decentralisation of data and/or distribution of control is either required or inherently present.
REINFORCEMENT LEARNING
While traditional Machine Learning (ML) techniques learn from examples (e.g. cat vs. dog), Reinforcement Learning (RL) learns from interaction with its environment. ML identifies patterns within a dataset, for example what characteristics of a photo make it a dog or a cat. Reinforcement learning, on the other hand, learns to perform a task, to take decisions, to optimize a reward given by a user – not unlike dog training when you give a cookie if the dog did well. It thus learns what the impact of your actions are on your environment. We call this a decision process.
In this context, we discriminate several important concepts:
The agent, who attempts to maximize its rewards by taking actions, in an “intelligent” way, to change its environment. An example environment could be a game of chess, with each action a potential move by the player, and the board state as the environment. Ultimately, there is one reward given at the end of the game, that is: have you won the game or not?
This agent thus observes current environmental states (e.g. through sensor readings or from camera images) and executes an action, such as moving or pressing a button. After every activity, the agent is given feedback in the form of a reward (a real number) which allows it to contemplate the next state.
Once a series of actions is executed or when the agent reaches a predefined goal, the “episode” finishes. The objective of the agent is to learn which action to execute in which state so that it obtains the maximum cumulative reward. In other words, the uttermost possible reward from the first observed state until the end of the episode. The agent is, then, reset, put back to a random environmental state, and starts executing actions again.
EVOLUTIONARY & HYBRID AI
Mission
We aim to build truly intelligent systems that are able to interact with and reason about their native environment in order to solve an open-ended set of tasks. Our systems are deeply inspired by evolutionary principles such as self-organisation, selection and emergent functionality, and are therefore adaptive by design. We adopt a hybrid approach that integrates symbolic and subsymbolic AI techniques, combining their strengths to achieve general, accurate and interpretable models. We focus in particular on tasks that require human language-like communication, involving advanced perception, reasoning and learning skills. We investigate fundamental research questions that have a tight connection to real-world problems.
Expertise
The EHAI team has extensive and unique expertise in the domains of emergent communication and computational construction grammar.
Emergent Communication
Which mechanisms does a population of agents need for constructing conceptual and linguistic structures that are adequate for interacting with, reasoning about and acting in their native environment?
Computational Construction Grammar
How can a machine learn the most basic function of natural language, namely that of mapping between rich meaning representations and expressive linguistic utterances?
EVOLUTIONARY LINGUISTICS
We investigate ways in which artificial agents can self-organize languages with natural-language like properties and how meaning can co-evolve with language. Our research is based on the hypothesis that language is a complex adaptive system that emerges through adaptive interactions between agents and continues to evolve in order to remain adapted to the needs and capabilities of the agents. We explore this hypothesis by implementing the full cycle of speaker and hearer as they play situated language games and observing the characteristics of the languages that emerge.
COMPUTATIONAL CREATIVITY
Computational creativity (CC) is the art, science, philosophy and engineering of computational systems which, by taking on particular responsibilities, exhibit behaviours that unbiased observers would deem to be creative. As a field of research, this area is thriving, with progress in formalising what it means for software to be creative, along with many exciting and valuable applications of creative software in the sciences, the arts, literature, gaming and elsewhere.
CC is a new way of looking at AI, which leaves behind the problem solving paradigm, and moves forward to problem-seeking approaches. In doing so, it echos some of the original motives of early AI, and also of the current Artificial General Intelligence (AGI) movement. The Association for Computational Creativity is the international organisation the promotes study of this new science.
At the VUB, Computational Creativity has a cognitive slant, with most of the work focused on predictive statistical systems that are claimed to model human creativity. One particular theory, Information Dynamics of Thinking, is at the centre of this work.
KNOWLEDGE REPRESENTATION & REASONING
Knowledge Representation and Reasoning (KRR) is a subfield of artificial intelligence (AI) concerned with defining artificial languages to express or represent knowledge and devising methodologies and tools to reason with this knowledge. Formal studies of knowledge representation and reasoning with knowledge date back to ancient times. As an example, consider Aristotle’s famous syllogism:
“All men are mortal.
Socrates is a man.
Therefore, Socrates is mortal.”
The argument presented above is inarguably valid. The three above sentences are now commonly represented in first-order logic (FO). While the Greeks relied on philosophers to make (formal) arguments like the one above, thanks to the progress in computer science, a simple personal computer can now automatically infer that the last sentence is indeed a consequence of the first two.
KRR is concerned with a broad variety of tasks. FO is only one of the many languages to represent knowledge in- and deduction, which is the kind of reasoning used above. Also, it is only one of many reasoning methods that are studied.
Some topics of interest of the KRR team of the AI lab include
- defining knowledge representation languages suitable for expressing certain types of knowledge;
- studying relationships between different languages
- defining inference methods (generic methods that allow exploiting the knowledge in a certain way)
- developing efficient algorithms for knowledge-based inference
- studying the relationship between knowledge representation and database theory
Facilities & Resources
Partner Organizations
Abbreviation |
VUB AI
|
Country |
Belgium
|
Region |
Europe
|
Primary Language |
French
|
Evidence of Intl Collaboration? |
|
Industry engagement required? |
Associated Funding Agencies |
Contact Name |
Ann Nowe
|
Contact Title |
Head of Lab
|
Contact E-Mail |
ann.nowe@vub.be
|
Website |
|
General E-mail |
info@ai.vub.ac.be
|
Phone |
32 (0)486 37.98.71
|
Address |
Pleinlaan 9
Brussels
1050
|
The Artificial Intelligence Lab, founded in 1983 by Prof. Dr. Luc Steels, is the first AI lab on the European mainland. It is headed by Prof. Dr. Ann Nowé and Prof. Dr. Bernard Manderick. During its history of more than three decades, the VUB AI Lab has been following two main routes towards the understanding of intelligence: both the symbolic route (classical AI) as the dynamics route (complex systems science). Academic track record Since its foundation, more than 50 people have received a PhD degree at the Artificial Intelligence Lab, of which 18 in the past 5 years only. The total amount of publications sums up to more than 850 publications which are cited more than 24.000 times, 8500 times in the last 5 years. The lab also has experience in setting up of spin-offs (5 since its beginning). The research group is provided with standard hardware and software utilities and departmental services (secretary and ICT support). The COMO group of the AI-lab focuses on the one hand on the modeling of natural phenomena, and on the other hand on developing algorithms for complex problem solving inspired by these natural phenomena. The lab has experience in a wide range of learning techniques such as: Reinforcement Learning, Genetic Algorithms, Neural Networks, Support Vector Machines, Graphical models including Bayesian Networks, Genetic algorithms etc. Industrial Cooperation Experience The AI lab has applied artificial intelligence techniques including machine learning in a wide range of settings in the past, from robotics (MIRAD – Vlaio SBO), Industry 4.0 (“OperatorInfo”) to epidemics (“Learning optimal preventive strategies to mitigate epidemics of latent infectious diseases” (FWO), in which individual based models are combined with multi-armed bandit problems to come with an ideal disease mitigation strategy. National and international network On the national level the lab collaborates with several research groups. For the proposed project the most relevant ones are: CIB KULeuven, Flanders Make (formerly FMTC), KULeuven Division of Production Engineering, Machine Design & Automation (PMA), INTEC UGent, Systems UGent, PATS UA, Applied Mathematics and Computer science UGent, Distrinet KULeuven, Mebios KULeuven. The lab is also member of the BruBotics Research Center, a VUB initiative which assembles interdisciplinary groups working on all aspects of robotics. The group also has a strong collaboration with their counterparts at the Université Libre de Bruxelles on topics related to Computational Biology and (Evolutionary) Game Theory. The group participates in a new Brussels institute for Bioinformatics research; the new institute called IB² gathers all medical and computational expertise in the Brussels area to tackle medical questions related to cancer, diabetes, etc. Through our alumni we have good connections with industry. They have taken jobs in AI companies like Google and DeepMind, and have been attracted by industry, often through contacts via SBO projects, to set up a team on data analytics (e.g Yann-Michael De Hauwere at Persgroep, Maarten Devillé at Medialaan and Maarten Peeters Forecasting Team Coordinator at ENGIE). There are also many international collaborations including: University of Minho and CFTC Lisbon, University of Maastricht (Kurt Driessens), University of York (Daniel Kudenko), Goldsmiths College London (Mark d’Inverno), Universität Wien (Tecumseh Fitch), University of Haifa (Wendy Sandler), Washington State University (Matt Taylor) and University of Liverpool (Karl Tuyls), and institutes in developing countries, in particular: University of Santa Clara and University of Camaguey in Cuba and University of Moi Kenya.
Abbreviation |
VUB AI
|
Country |
Belgium
|
Region |
Europe
|
Primary Language |
French
|
Evidence of Intl Collaboration? |
|
Industry engagement required? |
Associated Funding Agencies |
Contact Name |
Ann Nowe
|
Contact Title |
Head of Lab
|
Contact E-Mail |
ann.nowe@vub.be
|
Website |
|
General E-mail |
info@ai.vub.ac.be
|
Phone |
32 (0)486 37.98.71
|
Address |
Pleinlaan 9
Brussels
1050
|
Research Areas
LEARNING IN MULTI-AGENT SYSTEMS
Telecommunications, economics, mobile robots, traffic simulation, electricity grids/smart grids … are all examples of systems in which decentralisation of data and/or distribution of control is either required or inherently present.
REINFORCEMENT LEARNING
While traditional Machine Learning (ML) techniques learn from examples (e.g. cat vs. dog), Reinforcement Learning (RL) learns from interaction with its environment. ML identifies patterns within a dataset, for example what characteristics of a photo make it a dog or a cat. Reinforcement learning, on the other hand, learns to perform a task, to take decisions, to optimize a reward given by a user – not unlike dog training when you give a cookie if the dog did well. It thus learns what the impact of your actions are on your environment. We call this a decision process.
In this context, we discriminate several important concepts:
The agent, who attempts to maximize its rewards by taking actions, in an “intelligent” way, to change its environment. An example environment could be a game of chess, with each action a potential move by the player, and the board state as the environment. Ultimately, there is one reward given at the end of the game, that is: have you won the game or not?
This agent thus observes current environmental states (e.g. through sensor readings or from camera images) and executes an action, such as moving or pressing a button. After every activity, the agent is given feedback in the form of a reward (a real number) which allows it to contemplate the next state.
Once a series of actions is executed or when the agent reaches a predefined goal, the “episode” finishes. The objective of the agent is to learn which action to execute in which state so that it obtains the maximum cumulative reward. In other words, the uttermost possible reward from the first observed state until the end of the episode. The agent is, then, reset, put back to a random environmental state, and starts executing actions again.
EVOLUTIONARY & HYBRID AI
Mission
We aim to build truly intelligent systems that are able to interact with and reason about their native environment in order to solve an open-ended set of tasks. Our systems are deeply inspired by evolutionary principles such as self-organisation, selection and emergent functionality, and are therefore adaptive by design. We adopt a hybrid approach that integrates symbolic and subsymbolic AI techniques, combining their strengths to achieve general, accurate and interpretable models. We focus in particular on tasks that require human language-like communication, involving advanced perception, reasoning and learning skills. We investigate fundamental research questions that have a tight connection to real-world problems.
Expertise
The EHAI team has extensive and unique expertise in the domains of emergent communication and computational construction grammar.
Emergent Communication
Which mechanisms does a population of agents need for constructing conceptual and linguistic structures that are adequate for interacting with, reasoning about and acting in their native environment?
Computational Construction Grammar
How can a machine learn the most basic function of natural language, namely that of mapping between rich meaning representations and expressive linguistic utterances?
EVOLUTIONARY LINGUISTICS
We investigate ways in which artificial agents can self-organize languages with natural-language like properties and how meaning can co-evolve with language. Our research is based on the hypothesis that language is a complex adaptive system that emerges through adaptive interactions between agents and continues to evolve in order to remain adapted to the needs and capabilities of the agents. We explore this hypothesis by implementing the full cycle of speaker and hearer as they play situated language games and observing the characteristics of the languages that emerge.
COMPUTATIONAL CREATIVITY
Computational creativity (CC) is the art, science, philosophy and engineering of computational systems which, by taking on particular responsibilities, exhibit behaviours that unbiased observers would deem to be creative. As a field of research, this area is thriving, with progress in formalising what it means for software to be creative, along with many exciting and valuable applications of creative software in the sciences, the arts, literature, gaming and elsewhere.
CC is a new way of looking at AI, which leaves behind the problem solving paradigm, and moves forward to problem-seeking approaches. In doing so, it echos some of the original motives of early AI, and also of the current Artificial General Intelligence (AGI) movement. The Association for Computational Creativity is the international organisation the promotes study of this new science.
At the VUB, Computational Creativity has a cognitive slant, with most of the work focused on predictive statistical systems that are claimed to model human creativity. One particular theory, Information Dynamics of Thinking, is at the centre of this work.
KNOWLEDGE REPRESENTATION & REASONING
Knowledge Representation and Reasoning (KRR) is a subfield of artificial intelligence (AI) concerned with defining artificial languages to express or represent knowledge and devising methodologies and tools to reason with this knowledge. Formal studies of knowledge representation and reasoning with knowledge date back to ancient times. As an example, consider Aristotle’s famous syllogism:
“All men are mortal.
Socrates is a man.
Therefore, Socrates is mortal.”
The argument presented above is inarguably valid. The three above sentences are now commonly represented in first-order logic (FO). While the Greeks relied on philosophers to make (formal) arguments like the one above, thanks to the progress in computer science, a simple personal computer can now automatically infer that the last sentence is indeed a consequence of the first two.
KRR is concerned with a broad variety of tasks. FO is only one of the many languages to represent knowledge in- and deduction, which is the kind of reasoning used above. Also, it is only one of many reasoning methods that are studied.
Some topics of interest of the KRR team of the AI lab include
- defining knowledge representation languages suitable for expressing certain types of knowledge;
- studying relationships between different languages
- defining inference methods (generic methods that allow exploiting the knowledge in a certain way)
- developing efficient algorithms for knowledge-based inference
- studying the relationship between knowledge representation and database theory