Program
Monday 7/11
Tuesday 8/11
Wednesday 9/11
| Mechelen Centraal (+1) | Auditorium (+3) | Alcazar (+5) | Dijlezaal ( 0) | |
|---|---|---|---|---|
| 08:30–09:00 | Registration & Welcome coffee | |||
| 09:00–09:30 | ||||
| 09:30–10:00 | Predictive models | Time-series | Industry | |
| 10:00–10:30 | ||||
| 10:30–11:00 | ||||
| Coffee break & Poster session 1 & Demos | ||||
| 11:00–11:30 | ||||
| 11:30–12:00 | Demos | Classification | RL & MDP | Language & creativity |
| 12:00–12:30 | ||||
| 12:30–13:00 | Lunch (on your own) | |||
| 13:00–13:30 | ||||
| 13:30–14:00 | ||||
| 14:00–14:30 | Human+AI | Scheduling | Logic | |
| 14:30–15:00 | ||||
| 15:00–15:30 | Coffee break & Poster session 2 | |||
| 15:30–16:00 | ||||
| 16:00–16:30 | Welcome speech | |||
| 16:30–17:00 | Invited talk — Marcin Detyniecki | |||
| 17:00–17:30 | ||||
| 17:30–18:00 | Reception | |||
| 18:00–18:30 | ||||
| 18:30–19:00 |
Sessions
Session chair: Wannes Meert
| Gianluca Bontempi | Between accurate prediction and poor decision making: the AI/ML gap (position paper) |
| Yuko Kato, David Tax and Marco Loog | A view on model misspecification in uncertainty quantification |
| Wim Fyen, Wannes Meert, Bart Thijs and Koenraad Debackere | Investigation of the predictive power of the human capital of academic researchers on their financial performance using social network analysis and decision trees |
| Nicolas Dewolf, Bernard De Baets and Willem Waegeman | Valid prediction intervals for regression problems |
| Dimitrios Iliadis, Bernard De Baets and Willem Waegeman | DeepMTP: a Python-based deep learning framework for multi-target prediction |
Session chair: Toon Calders
| Wolf De Wulf | Transfer learning in Brain-Computer Interfaces: Language-Pretrained Transformers for Classifying Electroencephalography |
| Stijn Rotman and Boris Čule | Forecasting Electric Vehicle Supply Equipment Availability |
| Victoria Catalán Pastor, Elena Congeduti, Aleksander Czechowski and Frans A. Oliehoek | Overcoming Traffic Sensors Malfunctions with Deep Learning |
| Alejandro Morales-Hernández, Gonzalo Nápoles, Agnieszka Jastrzębska, Yamisleydi Salgueiro Sicilia and Koen Vanhoof | Online learning of windmill time series using Long Short-term Cognitive Networks |
| Azqa Nadeem and Sicco Verwer | SECLEDS: Sequence Clustering in Evolving Data Streams via Multiple Medoids and Medoid Voting (Encore abstract) |
Session chair: Sophia Katrenko
Session chair: David Tax
| Joseph Mietkiewicz and Anders Madsen | Improvement of the fine tuning algorithm for naïve Bayes |
| Gadzhi Musaev, Kevin Mets, Tom De Schepper, Peter Hellinckx, Rokas Tamošiūnas and Vadim Uvarov | On-Device Deep Learning Location Category Inference Model |
| Oliver Urs Lenz, Daniel Peralta and Chris Cornelis | Optimised one-class classification performance |
| Bojian Yin, Federico Corradi and Sander Bohte | Accurate and efficient time-domain classification with adaptive spiking recurrent neural networks |
Session chair: Johan Kwisthout
| Rolf Starre, Marco Loog and Frans Oliehoek | Model-Based Reinforcement Learning with State Abstraction: A Survey |
| Elena Congeduti and Frans Oliehoek | A Cross-Field Review of State Abstraction for Markov Decision Processes |
| Merlijn Krale, Thiago Simão and Nils Jansen | The Value of Measuring in Q-learning for Markov Decision Processes |
| Florent Delgrange, Ann Nowé and Guillermo A. Pérez | Distillation of RL Policies with Formal Guarantees via Variational Abstraction of Markov Decision Processes |
Session chair: Celine Vens
| Ethelbert Uzodinma and Jennifer Spenader | Improving Domain Robustness in Out-of-Domain Corpus |
| Cédric Goemaere, Thomas Demeester, Tim Verbelen, Bart Dhoedt and Cedric De Boom | Efficient Keyword Generation using Pretrained Language Models |
| Lin de Huybrecht, Nick Harley and Geraint Wiggins | Text Rendering for Automatic Story Generation |
| Luc Steels and Lara Verheyen | Quantifying the contribution of different knowledge sources in text understanding |
Session chair: Elias Fernández
| Bert Bredeweg and Marco Kragten | Requirements and challenges for hybrid intelligence: A case-study in education |
| Emre Erdogan, Frank Dignum, Rineke Verbrugge and Pinar Yolum | Computational Theory of Mind for Human-Agent Collaboration |
| Michiel van der Meer, Catholijn Jonker and Pradeep Kumar Murukannaiah | HyEnA: Hybrid Intelligence for Argument Mining |
| Abdo Abouelrous, Yingqian Zhang and Laurens Bliek | Digital Twin Applications in Urban Logistics: An Overview |
Session chair: Jefrey Lijffijt
| Seppe Renty, Raphaël Avalos, Andries Rosseau and Ann Nowé | Tackling Scheduling Problems With Graph Structured Reinforcement Learning. |
| Nikolaos Efthymiou and Neil Yorke-Smith | Machine Learning for the Cyclic Hoist Scheduling Problem |
| Menno Oudshoorn, Timo Koppenberg and Neil Yorke-Smith | Optimisation of Annual Planned Rail Maintenance (Article Abstract) |
| Issa Hanou, Mathijs de Weerdt and Jesse Mulderij | Applying the Pebble Motion problem: studying the feasibility of the Train Unit Shunting Problem |
Session chair: Hendrik Blockeel
| Nico Roos | Specificity and context dependent preferences in argumentation systems |
| Pierre Carbonnelle, Simon Vandevelde, Joost Vennekens and Marc Denecker | IDP-Z3: a reasoning engine for FO(.) |
| Johan Kwisthout | Expanding Bayesian networks |
| Floris Geerts, Jasper Steegmans and Jan Van den Bussche | On the expressive power of message-passing neural networks as global feature map transformers |
Posters & Demos
Posters of previous sessions:
| Marcio Fuckner, Suzana Bašić, Pascal Wiggers, Yuri Westplat, Rick van Kersbergen and Iman Firouzifard | BubbleMachine: An agent-based modelling tool to analyse the interplay between cognitive preferences, social network interactions and algorithmic curation |
| Matthias Müller-Brockhausen and Hélène Plisnier | Transferring While Playing the RL Agent |
| Jeff van de Kamer, Maaike Hovenkamp, Erik Puik and Diederik M. Roijers | Monitoring Diabetic Foot Ulceration Treatment with Smart Insoles and Neural Networks |
Posters of previous sessions:
Invited Talk
Auditorium 16:30–17:30
Session chair: Toon Calders
When responsible AI research needs to meet reality
Responsible AI topics are now firmly established in the research community, with explainable AI or AI fairness as flagships. On the other end of the spectrum, those topics are now seriously attracting attention in organizations and companies, especially through the lens of AI governance, a discipline that aims to analyze and address the challenges arising from a widespread use of AI in practice, as AI regulations are around the corner.
This leads companies to focus on the compliance aspects of AI projects. In practice, data scientists and organizations are in a fog, missing adequate guidance and solutions from research to achieve responsible AI.
In this talk, we will discuss how large companies, like AXA, currently see the responsible AI topic and why the current research output only provides partially actionable methodologies and solutions. We will discuss and illustrate with some concrete examples how the research community could better address the scientific challenges of this new applied responsible AI practice.
AXA, France
Short bio
| Mechelen Centraal (+1) | Auditorium (+3) | Alcazar (+5) | Dijlezaal ( 0) | |
|---|---|---|---|---|
| 08:30–09:00 | Registration & Welcome coffee | |||
| 09:00–09:30 | NLP | Meta-learning | Health applications | |
| 09:30–10:00 | ||||
| 10:00–10:30 | ||||
| 10:30–11:00 | Coffee break & Poster session 1 & Demos | |||
| 11:00–11:30 | ||||
| 11:30–12:00 | Demos | Classification | Reinforcement learning | Time-series |
| 12:00–12:30 | ||||
| 12:30–13:00 | Lunch (on your own) | |||
| 13:00–13:30 | ||||
| 13:30–14:00 | ||||
| 14:00–14:30 | Optimization | Multi-objective RL | Computational creativity | |
| 14:30–15:00 | ||||
| 15:00–15:30 | Coffee break & Poster session 2 | |||
| 15:30–16:00 | ||||
| 16:00–16:30 | FACt (FACulty focusing on the FACts of AI) | |||
| 16:30–17:00 | ||||
| 17:00–17:30 | Invited talk 2 — Tinne Tuytelaars | |||
| 17:30–18:00 | ||||
| 18:00–18:30 | ||||
| 18:30–19:00 | ||||
| 19:00–19:30 | Conference dinner & award ceremony | |||
| 19:30–20:00 | ||||
| …–22:00 |
Sessions
Session chair: Paul Van Eecke
| Jens Nevens, Jonas Doumen, Paul Van Eecke and Katrien Beuls | Language Acquisition through Intention Reading and Pattern Finding |
| Paul Keuren, Marc Ponsen and Ayoub Bagheri | WordGraph2Vec: Combining domain knowledge with embeddings |
| Jeska Buhmann, Maxime De Bruyn, Ehsan Lotfi and Walter Daelemans | Domain- and Task-Adaptation for VaccinChatNL, a Dutch COVID-19 FAQ Answering Corpus and Classification Model |
| Marlon Saelens, Marie-Francine Moens, Ruben Cartuyvels and Liesbeth Allein | Implicit Reasoning over Temporal Relations in Evidence-Based Fact-Checking |
Session chair: Gianluca Bontempi
| Ozgur Taylan Turan, David M.J. Tax and Marco Loog | When MAML Learns Quickly, Does It Generalize Well? |
| Zhiyi Chen, Marco Loog and Jesse H. Krijthe | Explaining Two Strange Learning Curves |
| Donghwi Kim and Tom Viering | Different approaches to fitting and extrapolating the learning curve |
| Felix Mohr, Tom Viering, Marco Loog and Jan van Rijn | LCDB 1.0: An Extensive Learning Curves Database for Classification Tasks |
| Prajit Bhaskaran and Tom Viering | To Tune or not to Tune: Hyperparameter Influence on the Learning Curve |
| Konstantinos Theodorakos, Mauricio Agudelo, Joachim Schreurs, Johan Suykens and Bart De Moor | Ozone forecasting across Belgium with co-evolutionary Neural Architecture Search |
Session chair: Isel Grau
| Arnon Vandenberghe, Lyse Naomi Wamba Momo, Vincent Scheltjens and Bart De Moor | Multimodal Deep Learning for Early Length of Stay Prediction using Patient Similarity Embeddings |
| Robbe Claeys, Yuju Ahn, Moobeom Hong, Inkyun Park and Jihwan Lim | Towards a Systematic Investigation of Deep Learning Approaches for Bacterial Taxonomic Classification Using the 16S rRNA Gene |
| Nele Albers, Mark A. Neerincx, Kristell M. Penfornis and Willem-Paul Brinkman | Using a Virtual Coach to Quit Smoking: 14 Themes for User Needs |
| Sander van Donkelaar, Lois Daamen, Paul Andel, Ralf Zoetekouw and Sharon Ong | Superpixel-based Context Restoration for Self-supervised Pancreas Segmentation from CT scans |
Session chair: Sharon Ong
| Yochem van Rosmalen, Florian van der Steen, Sebastiaan Jans and Daan van der Weijden | Bursting the Burden Bubble? An Assessment of Sharma et al.’s Counterfactual-Based Fairness Metric |
| Steven Michiels, Lynn Houthuys, Cédric De Schryver, Frederik Vogeler and Frederik Desplentere | Machine learning for automated quality control in injection moulding manufacturing |
| Rodi Laanen, Maedeh Nasri, Richard van Dijk, Mitra Baratchi, Alexander Koutamanis and Carolien Rieffe | Automated classification of pre-defined movement patterns: A comparison between GNSS and UWB technology |
| Afsana Khan, Marijn ten Thij and Anna Wilbik | Communication-Efficient Vertical Federated Learning |
Session chair: Diederik Roijers
| Astrid Vanneste, Thomas Somers, Simon Vanneste, Kevin Mets, Tom De Schepper, Siegfried Mercelis and Peter Hellinckx | Scalability of Message Encoding Techniques for Continuous Communication Learned with Multi-Agent Reinforcement Learning |
| Matthias Hutsebaut-Buysse, Kevin Mets, Tom De Schepper and Steven Latre | Structured Exploration Through Instruction Enhancement for Object Navigation |
| Grigory Neustroev, Sytze Andringa, Remco Verzijlbergh and Mathijs de Weerdt | Deep Reinforcement Learning for Active Wake Control |
| Santiago Amaya | Modelling of human behaviour in traffic interactions using Inverse Reinforcement Learning |
Session chair: Len Feremans
| Rickard K.A. Karlsson, Martin Willbo and Fredrik D. Johansson | Learning using Privileged Time-Series |
| Isel Grau, Michiel de Hoop, Ana Glaser, Gonzalo Nápoles and Remco Dijkman | Semiconductor Demand Forecasting using Long Short-term Cognitive Networks |
| Daan Van Wesenbeeck, Aras Yurtman, Wannes Meert and Hendrik Blockeel | Unsupervised extraction and clustering of physical therapy exercise executions |
| Jean Dessain | Improving the prediction of asset returns with machine learning by using a custom loss function |
Session chair: Bart Bogaerts
| Daimy Van Caudenberg and Bart Bogaerts | Static Symmetry and Dominance Breaking for Pseudo-Boolean Optimization |
| Ya Song, Laurens Bliek and Yingqian Zhang | Algorithm Selection for Traveling Salesman Problem with Simplified PointNet++ |
| Sho Cremers, Valentin Robu, Daan Hofman, Titus Naber, Kawin Zheng and Sonam Norbu | Efficient Methods for Approximating the Shapley Value for Asset Sharing in Energy Communities (Encore Abstract) |
| Dorian Vandenthoren | Nonnegative Bilinear Matrix Factorization |
Session chair: Pieter Libin
| Conor F Hayes, Timothy Verstraeten, Diederik M. Roijers, Enda Howley and Patrick Mannion | Expected scalarised returns dominance: a new solution concept for multi-objective decision making |
| Conor F Hayes, Roxana Radulescu, Eugenio Bargiacchi, Johan Källström, Matthew Macfarlane, Mathieu Reymond, Timothy Verstraeten, Luisa M Zintgraf, Richard Dazeley, Fredrik Heintz, Enda Howley, Athirai A. Irissappane, Patrick Mannion, Ann Nowé, Gabriel Ramos, Marcello Restelli, Peter Vamplew and Diederik M. Roijers | Abstract: A Practical Guide to Multi-Objective Reinforcement Learning and Planning |
| Willem Röpke, Diederik M. Roijers, Roxana Rădulescu and Ann Nowé | Communication In Multi-Objective Games |
Session chair: Joost Vennekens
| Deniz Ezgi Kurt and Egberdien van der Peijl | Is this art? A trans-disciplinary study of metadata in AI using the categorisation of ‘artists’ writing’ coined by the Art and Language Group (1968 – ). |
| Jeanine-Estelle Vallecalle | Overview of psychometric tools to evaluate robotic creativity – A scoping review |
| Marnix Verduyn | Comic Art Generation using GANs |
| Erinn Van der Sande, Nick Harley and Geraint Wiggins | Fabula Generation with BCDI Characters and Narrative Tension |
| Khalid Al Khatib (University of Groningen) | Deliberation, Not Persuasion: How AI Can Support More Effective Discussions |
| Joost Vennekens (KU Leuven) | Using Logic to Empower the User |
| Gianluca Bontempi (Université Libre de Bruxelles) | AI decision-making and ethics: what data scientists should know |
Posters & Demos
Posters of previous sessions:
| Willem Röpke, Samuele Pollaci, Bram Vandenbogaerde, Jiahong Li and Youri Coppens | Multi-Objective Scheduling for Agricultural Interventions |
| Daphne Lenders and Toon Calders | A New Benchmarking Dataset for Fair ML |
| Lucas N. Alegre, Florian Felten, El-Ghazali Talbi, Grégoire Danoy, Ann Nowé, Ana L. C. Bazzan and Bruno C. da Silva | MO-Gym: A Library of Multi-Objective Reinforcement Learning Environments |
Posters of previous sessions:
Invited Talk
Auditorium 17:00–18:00
Session chair: Celine Vens
Continual learning: Beyond solving datasets
KU Leuven, Belgium
Short bio
Tinne Tuytelaars is professor at KU Leuven, Belgium, working on computer vision and, in particular, topics related to image representations, vision and language, incremental learning, image generation and more. She has been program chair for ECCV14, general chair for CVPR16, and will again be program chair for CVPR21. She also served as associate-editor-in-chief of the IEEE Transactions on Pattern Analysis and Machine Intelligence over the last four years. She was awarded an ERC Starting Grant in 2009 and received the Koenderink test-of-time award at ECCV16.
| Mechelen Centraal (+1) | Auditorium (+3) | Alcazar (+5) | Dijlezaal ( 0) | |
|---|---|---|---|---|
| 08:30–09:00 | Registration & Welcome coffee | |||
| 09:00–09:30 | Explainable AI | Image classification and segmentation | Ontologies and knowledge graphs | |
| 09:30–10:00 | ||||
| 10:00–10:30 | ||||
| 10:30–11:00 | Coffee break & Poster session 1 & Demos | |||
| 11:00–11:30 | ||||
| 11:30–12:00 | Demos | Classification | Reinforcement learning | NLP |
| 12:00–12:30 | ||||
| 12:30–13:00 | Lunch provided | |||
| 13:00–13:30 | General assembly BNVKI | |||
| 13:30–14:00 | ||||
| 14:00–14:30 | Runner-up awards | Runner-up awards | Runner-up awards | |
| 14:30–15:00 | ||||
| Coffee break & Poster session 2 | ||||
| 15:00–15:30 | ||||
| 15:30–16:00 | ||||
| 16:00–16:30 | Best demo announcement & Best paper and thesis presentations | |||
| 16:30–17:00 | ||||
| 17:00–17:30 | Invited talk 3 — Karl Tuyls | |||
| 17:30–18:00 |
Sessions
Session chair: Jan Lemeire
| Arne Gevaert and Yvan Saeys | PDD-SHAP: Fast Approximations for Shapley Values using Functional Decomposition |
| Wai Wong, Walter Schaeken and Joost Vennekens | On the importance of experimental psychology for explainable artificial intelligence |
| Raphaela Butz, Arjen Hommersom, Marco Barenkamp and Hans van Ditmarsch | One counterfactual does not make an explanation |
| Roan Schellingerhout, Volodymyr Medentsiy and Maarten Marx | Explainable Career Path Predictions using Neural Models |
| Suzana Bašić, Marcio Fuckner and Pascal Wiggers | Explainable Misinformation Detection from Text: A Critical Look |
| Carla Wrede, Mark Winands and Anna Wilbik | Linguistic Summaries as Explanation Mechanism for Classification Problems |
Session chair: Jef Lijffijt
| Britt Schmitz and Marc Ponsen | Change Detection of Land Use: A Deep Learning Case-Study |
| Jef Plochaet and Toon Goedemé | Towards Automatic Proctoring of Online Exams Using Video Anomaly Detection |
| Arya Tri Prabawa, Merel Jung, Kostas Stoitsas, Werner Liebregts and Itır Önal Ertuğrul | Predicting Probability of Investment Based on Investor’s Facial Expression in a Startup Funding Pitch |
| Thomas Cassimon, Liam Hertoghs, Simon Vanneste, Phil Reiter, Kevin Mets, Tom De Schepper, Siegfried Mercelis and Peter Hellinckx | Predicting Image Classifier Performance Using the Synthetic Petri Dish Method |
| Benjamin Rombaut, Joris Roels and Yvan Saeys | BioSegment: Active Learning segmentation for 3D electron microscopy imaging |
| Mirela Popa, Marius Sommerfeld, Joël Karel, Stefan Schwartz and András Tóth | 3D Plant Segmentation for High Throughput Phenotyping using 3D Morphological Algorithms |
Session chair: Jan Van den Bussche
| Idries Nasim and Shuai Wang | Examining the Evolution of Identity graphs Through Redirection |
| Tianyang Lu, Shuai Wang and Zhisheng Huang | Towards Smart Urban Resilience: A Linked Data Approach |
| Romana Pernisch, Daniele Dell’Aglio, Mirko Serbak, Rafael Gonçalves and Abraham Bernstein | Visualising the Effects of Ontology Changes and Studying their Understanding with ChImp |
| Loan Ho, Somjit Arch-Int, Erman Acar, Stefan Schlobach and Ngamnij Arch-Int | An Argumentative Approach for Handling Inconsistency in Prioritized Datalog± Ontologies |
| Dimitrios Alivanistos, Max Berrendorf, Michael Cochez and Michael Galkin | Query Embedding on Hyper-Relational Knowledge Graphs |
Session chair: Lynn Houthuys
| Thi Hoang Anh Tran, Malina Lara Wiesner and Maurice van Keulen | Influence of discretization granularity on learning classification models |
| Daniël Vos and Sicco Verwer | Robust Optimal Classification Trees against Adversarial Examples |
| Tobias Glasmachers | Recipe for Fast Large-scale SVM Training: Polishing, Parallelism, and more RAM! |
| Thomas Mortier, Eyke Hüllermeier, Krzysztof Dembczyński and Willem Waegeman | Set-valued prediction in hierarchical classification |
Session chair: Kevin Mets
| Floris den Hengst, Vincent François-Lavet, Mark Hoogendoorn and Frank van Harmelen | Reinforcement Learning with Option Machines |
| Louis Bagot, Kevin Mets, Tom De Schepper and Steven Latré | A Case for Feature-Based Successor Features for Transfer in Reinforcement Learning |
| Ksenija Stepanovic, Jichen Wu, Rob Everhardt and Mathijs de Weerdt | Unlocking the Flexibility of District Heating Pipeline Energy Storage with Reinforcement Learning |
| Julius Wagenbach and Matthia Sabatelli | Factors of Influence of the Overestimation Bias of Q-Learning |
Session chair: Gerasimos (Jerry) Spanakis
| Rastislav Hronsky and Emmanuel Keuleers | Does the Choice of a Segmentation Algorithm Affect the Performance of Text Classifiers? |
| Sander Brinkhuijsen, Romana Pernisch, Eljo Haspels and Mark Van Staalduinen | Context-Aware Feature Vectors in Dark Web Page Classification |
| Yawen Zhao and Jordy Van Landeghem | Better Late than Never: Late Fusion Techniques for Document Classification |
| Daniel Daza, Michael Cochez and Paul Groth | SlotGAN: Detecting Mentions in Text via Adversarial Distant Learning |
Runner-up Award Nominees
Auditorium
Auditorium 14:00–14:45
Session chair: Toon Calders
| Liesbet De Vos | Human-Interpretable Grounded Language Processing |
| Salima Lamsiyah and Christoph Schommer | A Comparative Study of Sentence Embeddings for Unsupervised Extractive Multi-Document Summarization |
Alcazar
Alcazar 14:00–14:45
Session chair: Celine Vens
| Shashank Subramanya and Gerasimos Spanakis | MultiTM: A Multilingual Topic Modeling approach based on Clustering |
| Paul Wallbott, Sascha Grollmisch and Thomas Köllmer | Examining speaker and keyword uniqueness: Partitioning keyword spotting datasets for federated learning with the largest differencing method |
Dijlezaal
Dijlezaal 14:00–14:45
Session chair: Jef Lijffijt
| William Dumez, Simon Vandevelde and Joost Vennekens | Step-wise Explanations of Sudokus using IDP |
| Marco Loog and Tom Viering | A Survey of Learning Curves with Bad Behavior: or How More Data Need Not Lead to Better Performance |
| Fieke Middelraad | This laptop has great coffee: Training a Dutch ABSA model from customer reviews |
| Kshitij Goyal, Wannes Meert, Hendrik Blockeel, Elia Van Wolputte, Koen Vanderstraeten, Wouter Pijpops and Kurt Jaspers | Automatic Generation of Product Concepts from Positive Examples, with an Application to Music Streaming |
Posters & Demos
Posters of previous sessions:
| Simon Vandevelde and Joost Vennekens | FOLL-E: Teaching First Order Logic to Children |
| Tias Guns, Milan Pesa, Maxime Mulamba, Ignace Bleukx, Emilio Gamba and Senne Berden | Sudoku Assistant – An AI-powered app to help solve pen-and-paper Sudokus |
| Ruben Janssens, Thomas Demeester and Tony Belpaeme | Visual Conversation Starters for Human-Robot Interaction |
| Axel Abels, Elias Fernández Domingos, Tom Lenaerts, Vito Trianni and Ann Nowé | Bias Mitigation in Decision-Making with Expert Advice |
Posters of previous sessions:
Invited Talk
Auditorium 17:00–18:00
Session chair: Jefrey Lijffijt
Mastering the Game of Stratego with Model-Free Multiagent Reinforcement Learning
In this talk I will introduce DeepNash, an autonomous agent capable of learning to play the imperfect information game Stratego from scratch, up to a human expert level. Stratego is one of the few iconic board games that Artificial Intelligence (AI) had not yet mastered. This game has an enormous game tree, orders of magnitude bigger than that of Go and Texas hold’em poker. It has the additional complexity of requiring decision-making under imperfect information, similar to Texas hold’em poker. Decisions in Stratego are made over a large number of discrete actions with no obvious link between action and outcome. Episodes are long, with often hundreds of moves before a player wins, and situations in Stratego can not easily be broken down into manageable-sized sub-problems as in poker. For these reasons, Stratego has been a grand challenge for the field of AI for decades, and existing AI methods barely reach an amateur level of play. DeepNash uses a game-theoretic, model-free deep reinforcement learning method, without search, that learns to master Stratego via self-play and converges to an approximate Nash equilibrium, instead of ‘cycling’ around it, by directly modifying the underlying multi-agent learning dynamics. DeepNash beats existing state-of-the-art AI methods in Stratego and achieved a yearly and all-time top-3 rank on the Gravon games platform, competing with human expert players.
DeepMind, France
University of Liverpool, UK