The large variety of heuristic and metaheuristic algorithms for hard optimization problems raises numerous interesting and challenging issues. Practitioners are confronted with the burden of selecting the most appropriate method, in many cases through an expensive algorithm configuration and parameter tuning process, and subject to a steep learning curve. Scientists seek theoretical insights and demand a sound experimental methodology for evaluating algorithms and assessing strengths and weaknesses. A necessary prerequisite for this effort is a clear separation between the algorithm and the experimenter, who, in too many cases, is "in the loop" as a crucial intelligent learning component. Both issues are related to designing and engineering ways of "learning" about the performance of different techniques, and ways of using past experience about the algorithm behavior to improve performance in the future. Intelligent learning schemes for mining the knowledge obtained from different runs or during a single run can improve the algorithm development and design process and simplify the applications of high-performance optimization methods. Combinations of algorithms can further improve the robustness and performance of the individual components provided that sufficient knowledge of the relationship between problem instance characteristics and algorithm performance is obtained.
Topics of Interest:
To recognize the value of interacting with outstanding experts, LION Extended Talks traditionally consist of two parts: a tutorial introduction followed by a state-of-the-art segment with a Q&A session
Abstract: TBD
Bio: The New York Times headlined: "When A.I. Matures, It May Call Jürgen Schmidhuber 'Dad'.” He is often called the Father of Modern AI by the media. In 1990-91, he laid foundations of "Generative AI," by introducing the principles of Generative Adversarial Networks (now used for deepfakes), unnormalised linear Transformers (see the T in ChatGPT), self-supervised Pre-Training for deep learning (see the P in ChatGPT), and network distillation (essential for the famous DeepSeek).
This early work on the "G", "P" and "T" in GPT has had a wide-ranging impact, and earned him the nickname "Father of Generative AI." His lab also produced LSTM, the most cited AI of the 20th century, and the Highway Net (a variant of which is the most cited AI of the 21st century). He also pioneered meta-learning machines that learn to learn (1987-), and neural AIs that set themselves their own goals (1990-). His formal theory of creativity & curiosity & fun (2006-2010) explains art, science, music, and humor. He also generalized algorithmic information theory and the many-worlds theory of physics (1997-2000). Elon Musk tweeted: "Schmidhuber invented everything." His AI is on over 3 billion smartphones, and used many billions of times per day.
Abstract: Mixed-Integer Programming (MIP) technology is used daily to solve (discrete) optimization problems in contexts as diverse as energy, transportation, logistics, telecommunications, biology, just to mention a few. The MIP roots date back to 1958 with the seminal work by Ralph Gomory on cutting plane generation. In this talk, we will discuss — taking the (biased) viewpoint of the speaker — how MIP evolved in its main algorithmic ingredients, namely preprocessing, branching, cutting planes and primal heuristics, to become a mature research field whose advances rapidly translate into professional, widely available software tools. We will then discuss the next phase of this process, where Artificial Intelligence and, specifically, Machine Learning are already playing a significant role, a role that is likely to increase.
Bio: Andrea Lodi is the Andrew H. and Ann R. Tisch Professor of operations research and information engineering at Cornell Tech, the Jacobs Technion-Cornell Institute, and Cornell Engineering.
Before joining Cornell, Lodi was a Herman Goldstine Fellow at the IBM Mathematical Sciences Department of New York and a full professor of operations research at the Department of Electrical, Electronic, and Information Engineering at the University of Bologna. He was also the Canada Excellence Research Chair in “Data Science for Real-time Decision Making” at Polytechnique Montréal.
Abstract: Markov decision processes (MDPs) constitute one of the predominant modeling and solution paradigms for dynamic decision problems affected by uncertainty. MDPs model the dynamics of a system through a random state evolution that generates rewards over time. The decision maker aims to select actions that influence this state evolution so as to maximize rewards.
In this talk, we review recent advances in MDPs along two directions: (i) the construction of data-driven policies that combine the (traditionally separated) tasks of estimating the system’s behavior and selecting actions that maximize rewards in the estimated system, and (ii) the exploitation of structure to solve large-scale problems. In view of (i), we will show how the consideration of data-driven policies naturally leads to the study of robust MDPs, where the decision maker combats overfitting by hedging against the worst system dynamics that are plausible under some given training data. We will also discuss how alternative models of robustness offer different trade-offs between the competing goals of out-of-sample performance and complexity of the involved policies and computations. As for (ii), we will review two types of structure that allow us to alleviate the well-known curse of dimensionality: weakly coupled MDPs that combine a potentially large number of MDPs via a small number of linking constraints, and factored MDPs whose states are represented by assignments of values to state variables that evolve and contribute to the system's rewards largely independently.Bio: Wolfram Wiesemann is Professor of Analytics & Operations at the Analytics, Marketing & Operations department at Imperial College Business School. His research interests evolve around decision-making under uncertainty, with applications to logistics, supply chain management and healthcare. Wolfram has served as an elected member of the boards of the Mathematical Optimization Society and the Stochastic Programming Society, and he currently serves as editor in-chief of Operations Research Letters as well as a department co-editor for Management Science.
All deadlines are Anywhere on Earth (AoE = UTC-12h).
Types of Submissions
When submitting a paper to LION 20, authors are required to select one of the following three types of papers:
You can submit original and unpublished work either as a long paper (12-15 pages, including references) or short paper (6-11 pages, including references). You can choose to add an appendix.
Paper Format
Please prepare your paper in English using the Lecture Notes in Computer Science (LNCS) template, which is available [ Here ]. Papers must be submitted in PDF.
Submission System
All papers must be submitted using OpenReview (link coming soon)
All abstracts (for presentation only) must be submitted using OpenReview (link coming soon)
Review Process
Please note that the review process for LION 20 is single blind.
Proceedings
Papers accepted into the LION20 proceedings will be published in Lecture Notes in Computer Science (LNCS).
Special Issue
Accepted papers at LION20 will be invited to submit extended versions to a special issue. All manuscripts are subject to peer review. The details of the SI and submissions will be available during LION20.
In addition to submissions about general LION themes, we also welcome submissions related to one of our special sessions. The special sessions will be part of the regular conference and are subject to the same peer-review as all other submissions. Please address proposals for special sessions to the TPC Chair: prof. Maximilian Schiffer - schiffer(AT)tum.de, with CC: roberto.battiti(AT)unitn.it
Organizers: Lin Xie^1, Yingqian Zhang ^2, and Yaoxin Wu^2 1 Chair of Information Systems and Business Analytics, Brandenburg University of Technology, Germany (email: lin.xie@b-tu.de) 2 Eindhoven University of Technology, Netherlands
Abstract: Large Language Models (LLMs) are emerging as powerful tools for optimization, complementing and extending traditional methods. Their ability to understand problem descriptions, generate heuristics, and support human-in-the-loop decision making opens new possibilities for modeling and solving complex real-world problems. This session invites contributions on how LLMs and related AI techniques can be applied to domains such as logistics, scheduling, energy systems, and finance. Topics of interest include AI-assisted problem formulation, heuristic discovery, end-to-end methods, hybrid approaches that combine LLMs with classical algorithms, and practical applications that demonstrate their impact. Special attention will be given to challenges of scalability, interpretability, and reliability in operational settings. The session aims to bring together researchers and practitioners to explore how LLMs can transform optimization research and practice.
Organizers:
Vittorio Maniezzo, Department of Computer Science, Univ. of Bologna, Italy, vittorio.maniezzo@unibo.it
Roberto Montemanni, Dept. of Sciences and Methods for Engineering, Univ. of Modena and Reggio Emilia, Italy - roberto.montemanni@unimore.it
Abstract:
Matheuristics — methods that leverage mathematical programming in heuristic contexts — have emerged as a powerful approach for solving complex optimization problems across a broad range of domains. As learning techniques and optimization models converge, new opportunities arise for approaches that combine solution quality, robustness, and scalability.
This special session brings together researchers and practitioners from operations research, machine learning, and artificial intelligence to discuss advancements, challenges, and applications of matheuristics in the era of data-driven decision-making. We invite contributions that either advance matheuristic methodologies or demonstrate their impact in practical settings, with the aim of bridging mathematical rigor and heuristic flexibility.
Topics of interest include novel hybrid designs, learning-enhanced matheuristics, decomposition and surrogate-based strategies, benchmarking studies and applications in logistics, energy, finance, and other domains.
Organizers:
Konstantinos Chatzilygeroudis, Laboratory of Automation and Robotics (LAR), University of Patras, Greece - costashatz@upatras.gr
Konstantinos Asimakopoulos, Laboratory of Automation and Robotics (LAR), University of Patras, Greece
Abstract: Can optimizers themselves be learned? Traditional solvers must re-solve from scratch every time even within families of very familiar problems. Typically the user needs to hand-design heuristics such as step sizes, branching rules etc. Another interesting way to approach this is to use reinforcement learning (RL) to learn decision rules that guide an optimization process across families of problem instances. In this special session, we invite research where RL discovers update rules, search heuristics, or branching strategies that outperform hand-crafted counterparts on families of problems.
Organizers: Carolina Crespi, Department of Mathematics and Computer Science, University of Catania – carolina.crespi@unict.it Alessio Mezzina, Department of Mathematics and Computer Science, University of Catania – alessio.mezzina@phd.unict.it Mario Pavone, Department of Mathematics and Computer Science, University of Catania – mpavone@dmi.unict.it
Abstract: Dynamic autonomous navigation is a challenging problem at the intersection of optimization, machine learning, and artificial intelligence. Real-world applications, such as robotics, autonomous vehicles, and, in general, exploration of unknown or evolving environments require intelligent agents to adapt in real time to incomplete information, uncertain dynamics, and limited resources. These problems naturally fall under the scope of hard optimization, where classical methods often need to be extended or combined with learning strategies to achieve robust and scalable solutions. This special session focuses on novel models, algorithms, and learning-based approaches to address uncertainty and complexity in autonomous navigation. It welcomes contributions on advanced optimization techniques, machine learning methods for decision-making, and hybrid systems that combine both. The session encourages work that emphasizes experimental methodologies, performance evaluation, and the application of results to real-world case studies. Covering both theoretical developments and practical implementations, it aims to advance the understanding of how optimization and learning can eDectively address the challenges of dynamic, uncertain, and multi-agent navigation scenarios.
Keywords: Optimization under uncertainty, Machine learning for optimization, Dynamic autonomous navigation, Real-time decision-making, Multi-agent systems
Topics of Interest: Relevant topics include, but are not limited to: Learning-based optimization methods for uncertain and dynamic environments, Adaptive and real-time decision-making algorithms, Hybrid approaches combining optimization and machine learning, Reinforcement learning and probabilistic models for navigation tasks, Multi-agent coordination and collective intelligence, Performance evaluation and algorithm selection for complex tasks, Robust and scalable techniques for hard optimization problems, Applications in robotics, transportation, disaster response, and smart infrastructures
Organizers: Om Prakash Vyas, International Institute of Information Technology Naya Raipur Chhattisgarh, India - director@iiitnr.edu.in Ramakrishna Bandi, International Institute of Information Technology Naya Raipur Chhattisgarh, India - ramakrishna@iiitnr.edu.in Vijaya J, International Institute of Information Technology Naya Raipur Chhattisgarh, India - vijaya@iiitnr.edu.in
Abstract: The growing availability of rich event log data has opened novel opportunities for combining data- driven optimization, artificial intelligence, and business process mining to achieve breakthrough improvements in operational performance and decision-making. Process Mining has grown from a diagnostic tool into a data-driven optimization framework combining Process Science, Data Science, and AI. Modern systems now use simple and effective optimization techniques such as swarm intelligence, evolutionary algorithms, mathematical optimization, and reinforcement learning to repair models, allocate resources, and remove bottlenecks. With digital twins, prescriptive analytics, and cloud platforms, continuous process simulation and improvement have become easier and more accessible. This special session focuses on the latest theoretical, methodological, and applied contributions at this intersection. Even with progress in discovery, conformance checking, and model enhancement, several research problems remain, including scalability, noisy or incomplete logs, multi-source data, and real-time decision needs. Optimization methods can address these issues by supporting multi-objective discovery, explainable conformance repair, real-time recommendations, and AI-assisted simulation. Using optimization in these areas helps create smarter, faster, and more reliable Process Mining solutions for modern organizations.
Keywords: Process mining
Topics of Interest: The optimization paradigm in the process mining context is being explored at following levels:
Organizers: Hossein Moosaei, Department of Informatics, Jan Evangelista Purkyně University, Czech Republic - hmoosaei@gmail.com Zbyšek Posel, Department of Informatics, Jan Evangelista Purkyně University, Czech Republic Paolo Giorgini, Department of Information Engineering and Computer Science, University in Trento, Italy - paolo.giorgini@unitn.it
Abstract: Digital transformation in healthcare requires intelligent optimization and learning-based methods capable of handling complex, uncertain, and data-intensive environments. The growing availability of clinical, biological, and imaging data presents new opportunities for developing AI-driven tools to support diagnostics, treatment planning, and personalized rehabilitation. However, challenges remain in ensuring data reliability, interoperability, and the scalability of optimization techniques in real clinical settings. This special session aims to bring together researchers working at the intersection of artificial intelligence, optimization, and digital health to discuss advances in algorithms, models, and applications that enhance decision-making in healthcare systems. Topics of interest include optimization-based diagnostic models, machine learning for medical data analysis, predictive and prescriptive healthcare analytics, intelligent planning of medical resources, and AI- assisted rehabilitation systems. The session particularly welcomes hybrid approaches combining learning and optimization to enable efficient, interpretable, and human-centered solutions for real-world healthcare problems.
Organizer: Prof. Absalom El-Shamir Ezugwu (organizer), North-West University, South Africa - Absalom.ezugwu@nwu.ac.za
PC Members: Prof. Nomali Ngobese, North-West University, South Africa Dr. Diego Oliva, Universidad de Guadalajara, Prof. Liyana Shuib, University of Malaya, Malaysia Dr. Saptadeep Biswas, National Institute of Technology Agartala, India Dr. Olaide Oyelade, North Carolina A&T University, USA Prof. Seyedali Mirjalili, Torrens University Australia, Australia Prof. Amir H. Gandomi, University of Technology Sydney, Australia Dr. Andronicus Akinyelu Ayobami, University of KwaZulu-Natal, South Africa Prof. Samarjit Kar, National Institute of Technology, Durgapur, India Dr. Mario A. Navarro, Universidad de Guadalajara, Mexico Dr. Seyed Jalaleddin Mousavirad. Mid Sweden University, Sweden Prof. Erik Cuevas, Universidad de Guadalajara, Mexico Prof. Oluwasefunmi Arogundade, Federal University of Agriculture, Abeokuta, Nigeria Dr. Sani Isah Abba, Prince Mohammad Bin Fahd University, Saudi Arabia
Abstract:
The global agricultural sector must reconcile increasing production demands with the
imperative to reduce its environmental impact. Achieving this requires a transition from
static, heuristic-based methods to closed-loop systems capable of data-driven perception,
adaptive control, and sequential decision-making. This session focuses on the integration of
machine learning for state estimation or situational awareness, as well as intelligent
optimisation for action planning, to enable autonomous, next-generation agricultural systems.
Specifically, the special session aims to investigate how learning-enabled models can
perceive agricultural environments using soil sensors, irrigation monitors, and remote
sensing, and how intelligent optimisation algorithms can subsequently act upon this data to
make autonomous, high-value decisions. The core of this session will be the tight feedback
loop between learning and optimisation, from the field level to the supply chain. We will
highlight contributions where machine learning models for yield prediction, soil property
estimation, or pest forecasting directly inform and are informed by optimisation frameworks
for resource allocation, scheduling, and control.
We welcome research that demonstrates this synergy across a spectrum of methodologies,
including Bayesian optimisation and GenAI for managing complex agricultural simulations,
reinforcement learning for adaptive closed-loop control, and multi-objective optimisation for
navigating the trade-offs between productivity, profitability, and sustainability.
Topics of Interest: We invite contributions demonstrating the synergistic integration of learning and intelligent optimisation, including but not limited to:
Organizers: Antonio Candelieri, Prof., Dept. Economics Management and Statistics, University of Milano-Bicocca, Italy - antonio.candelieri@unimib.it
Abstract: Learning-and-optimization through Bayesian Optimization (BO) has allowed to successfully solve many real- life and industrial-relevant problems, while calling for extensions of the vanilla framework and new ways to investigate the role of surrogate model and acquisition function, as well as their interplay. This Special Session will consider papers addressing both relevant real-life applications entailing exotic BO – such as multi/many objective BO (also with preference learning), grey-box BO, multi-fidelity and multiple information source BO, constrained BO, and high-dimensional BO – and emerging challenges – such as the interplay between BO and Large Language Models, convergence guarantees under model-misspecification, human-assisted BO (aka human-in-the-loop BO), adaptive/dynamic exploration-exploitation trade-off mechanisms.
Best Paper Awards
The Best Paper Awards at LION20 will recognize outstanding contributions based on originality, technical quality, and impact. Selected papers will be honored during the conference, with each recipient receiving a certificate and a monetary prize.
Francesco Archetti
(Consorzio Milano Ricerche, Italy)
Antonio Candelieri,
University of Milano-Bicocca, Italy
Maximilian
Schiffer
(Professor for Business Analytics & Intelligent Systems, Technical University of Munich,
Germany).
Roberto
Battiti
(University of Trento, Italy - Head of the Steering Committee)
Francesco
Archetti
(Consorzio Milano Ricerche, Italy)
Christian
Blum (Spanish
National Research Council (CSIC), Spain)
Mauro
Brunato (University
of Trento, Italy)
Carlos A.
Coello-Coello
(CINVESTAV-IPN, Mexico)
Clarisse
Dhaenens
(University of Lille, France)
Paola Festa
(University of
Napoli, Italy)
Martin
Charles Golumbic
(University of Haifa, Israel)
Youssef
Hamadi (Tempero
Tech, France)
Milan
Hladik, Charles
University, Prague, Czech Republic
Laetitia
Jourdan
(University of Lille, France)
Nikolaos Matsatsinis
(Technical University of Crete, Greece)
Hossein
Moosaei, Jan
Evangelista Purkyně University, Czech Republic
Panos
Pardalos
(University of Florida, USA)
Mauricio
Resende
(University of Washington, USA)
Meinolf
Sellmann
(InsideOpt, USA)
Yaroslav
Sergeyev
(University of Calabria, Italy)
Dimitris
Simos (SBA
Research, Austria)
Thomas
Stuetzle
(University of Bruxelles, Belgium)
Kevin
Tierney (Bielefeld
University, Germany)
Yingqian Zhang,
Eindhoven University of Technology, The Netherlands.
The LION 20 will be hosted by Universita' degli Studi Milano Bicocca and Grand Hotel Villa Torretta and in the Universita' degli Studi Milano Bicocca. The location is connected to the historic center via subway (about 30 mins) and easily reachable from the local airports: Malpensa, Orio al Serio (aka "Milan-Bergamo") and Linate. Reserve here at Villa Torretta. Room availability at Villa Torretta is limited, but there are many other hotels and bed and breakfasts nearby. If you are willing to travel 20 minutes by Metro the offer is very large but be sure to reserve in advance, Milano can get very busy in certain periods. Booking.com (around University of Milano Bicocca) Airbnb.com (around Villa Torretta)
| Conference fees | ||
|---|---|---|
| Early Registration | Late Registration | |
| Regular | Eur 520 | Eur 620 |
| Student | Eur 450 | Eur 520 |
| Accompanying Person | Eur 200 | Eur 200 |
Fees include: Participation to all sessions, conference materials, publication
of accepted
papers in LNCS, LION association, coffee breaks, lunches, welcome meeting, conference dinner, and social
program.
Accompanying person fee includes: Conference dinner and Social program.
Please contact us by email regarding any queries you may have in relation to the conference or general information.Email: antonio.candelieri@unimib.it