Author name cluster

Giorgos Stamou

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

9 papers

2 author rows

ICML Conference 2025 Conference Paper

Enhancing Cooperative Multi-Agent Reinforcement Learning with State Modelling and Adversarial Exploration

Andreas Kontogiannis
Konstantinos Papathanasiou
Yi Shen 0011
Giorgos Stamou
Michael M. Zavlanos
George A. Vouros

Learning to cooperate in distributed partially observable environments with no communication abilities poses significant challenges for multi-agent deep reinforcement learning (MARL). This paper addresses key concerns in this domain, focusing on inferring state representations from individual agent observations and leveraging these representations to enhance agents’ exploration and collaborative task execution policies. To this end, we propose a novel state modelling framework for cooperative MARL, where agents infer meaningful belief representations of the non-observable state, with respect to optimizing their own policies, while filtering redundant and less informative joint state information. Building upon this framework, we propose the MARL SMPE$^2$ algorithm. In SMPE$^2$, agents enhance their own policy’s discriminative abilities under partial observability, explicitly by incorporating their beliefs into the policy network, and implicitly by adopting an adversarial type of exploration policies which encourages agents to discover novel, high-value states while improving the discriminative abilities of others. Experimentally, we show that SMPE$^2$ outperforms a plethora of state-of-the-art MARL algorithms in complex fully cooperative tasks from the MPE, LBF, and RWARE benchmarks.

Details

ICML Conference 2025 Conference Paper

SCENIR: Visual Semantic Clarity through Unsupervised Scene Graph Retrieval

Nikolaos Chaidos
Angeliki Dimitriou
Maria Lymperaiou
Giorgos Stamou

Despite the dominance of convolutional and transformer-based architectures in image-to-image retrieval, these models are prone to biases arising from low-level visual features, such as color. Recognizing the lack of semantic understanding as a key limitation, we propose a novel scene graph-based retrieval framework that emphasizes semantic content over superficial image characteristics. Prior approaches to scene graph retrieval predominantly rely on supervised Graph Neural Networks (GNNs), which require ground truth graph pairs driven from image captions. However, the inconsistency of caption-based supervision stemming from variable text encodings undermine retrieval reliability. To address these, we present SCENIR, a Graph Autoencoder-based unsupervised retrieval framework, which eliminates the dependence on labeled training data. Our model demonstrates superior performance across metrics and runtime efficiency, outperforming existing vision-based, multimodal, and supervised GNN approaches. We further advocate for Graph Edit Distance (GED) as a deterministic and robust ground truth measure for scene graph similarity, replacing the inconsistent caption-based alternatives for the first time in image-to-image retrieval evaluation. Finally, we validate the generalizability of our method by applying it to unannotated datasets via automated scene graph generation, while substantially contributing in advancing state-of-the-art in counterfactual image retrieval. The source code is available at https: //github. com/nickhaidos/scenir-icml2025.

Details

NeurIPS Conference 2025 Conference Paper

V-CECE: Visual Counterfactual Explanations via Conceptual Edits

Nikolaos Spanos
Maria Lymperaiou
Giorgos Filandrianos
Konstantinos Thomas
Athanasios Voulodimos
Giorgos Stamou

Recent black-box counterfactual generation frameworks fail to take into account the semantic content of the proposed edits, while relying heavily on training to guide the generation process. We propose a novel, plug-and-play black-box counterfactual generation framework, which suggests step-by-step edits based on theoretical guarantees of optimal edits to produce human-level counterfactual explanations with zero training. Our framework utilizes a pre-trained image editing diffusion model, and operates without access to the internals of the classifier, leading to an explainable counterfactual generation process. Throughout our experimentation, we showcase the explanatory gap between human reasoning and neural model behavior by utilizing both Convolutional Neural Network (CNN), Vision Transformer (ViT) and Large Vision Language Model (LVLM) classifiers, substantiated through a comprehensive human evaluation.

PDF Details

ICML Conference 2024 Conference Paper

Structure Your Data: Towards Semantic Graph Counterfactuals

Angeliki Dimitriou
Maria Lymperaiou
Giorgos Filandrianos
Konstantinos Thomas
Giorgos Stamou

Counterfactual explanations (CEs) based on concepts are explanations that consider alternative scenarios to understand which high-level semantic features contributed to particular model predictions. In this work, we propose CEs based on the semantic graphs accompanying input data to achieve more descriptive, accurate, and human-aligned explanations. Building upon state-of-the-art (SotA) conceptual attempts, we adopt a model-agnostic edit-based approach and introduce leveraging GNNs for efficient Graph Edit Distance (GED) computation. With a focus on the visual domain, we represent images as scene graphs and obtain their GNN embeddings to bypass solving the NP-hard graph similarity problem for all input pairs, an integral part of CE computation process. We apply our method to benchmark and real-world datasets with varying difficulty and availability of semantic annotations. Testing on diverse classifiers, we find that our CEs outperform previous SotA explanation models based on semantics, including both white and black-box as well as conceptual and pixel-level approaches. Their superiority is proven quantitatively and qualitatively, as validated by human subjects, highlighting the significance of leveraging semantic edges in the presence of intricate relationships. Our model-agnostic graph-based approach is widely applicable and easily extensible, producing actionable explanations across different contexts. The code is available at https: //github. com/aggeliki-dimitriou/SGCE.

Details

IJCAI Conference 2023 Conference Paper

Choose your Data Wisely: A Framework for Semantic Counterfactuals

Edmund Dervakos
Konstantinos Thomas
Giorgos Filandrianos
Giorgos Stamou

Counterfactual explanations have been argued to be one of the most intuitive forms of explanation. They are typically defined as a minimal set of edits on a given data sample that, when applied, changes the output of a model on that sample. However, a minimal set of edits is not always clear and understandable to an end-user, as it could constitute an adversarial example (which is indistinguishable from the original data sample to an end-user). Instead, there are recent ideas that the notion of minimality in the context of counterfactuals should refer to the semantics of the data sample, and not to the feature space. In this work, we build on these ideas, and propose a framework that provides counterfactual explanations in terms of knowledge graphs. We provide an algorithm for computing such explanations (given some assumptions about the underlying knowledge), and quantitatively evaluate the framework with a user study.

PDF Details DOI

IJCAI Conference 2016 Conference Paper

Efficient Query Answering over Expressive Inconsistent Description Logics

Eleni Tsalapati
Giorgos Stoilos
Giorgos Stamou
George Koletsos

Inconsistent-tolerant semantics, like the IAR and ICAR semantics, have been proposed as means to compute meaningful query answers over inconsistent Description Logic (DL) ontologies. In the current paper we present a framework for scalable query answering under both the IAR and ICAR semantics, which is based on highly efficient data saturation systems. Our approach is sound and complete for ontologies expressed in the lightweight DL DL-Lite, but for more expressive DLs the problem is known to be intractable, hence our algorithm only computes upper approximations. Nevertheless, its structure motivates a new type of ICAR-like semantics which can be computed in polynomial time for a very large family of DLs. We have implemented our techniques and conducted an experimental evaluation obtaining encouraging results as both our IAR- and ICAR-answering approaches are far more efficient thanexisting available IAR-based answering systems.

PDF Details

AAAI Conference 2015 Conference Paper

Lower and Upper Bounds for SPARQL Queries over OWL Ontologies

Birte Glimm
Yevgeny Kazakov
Ilianna Kollia
Giorgos Stamou

The paper presents an approach for optimizing the evaluation of SPARQL queries over OWL ontologies using SPARQL’s OWL Direct Semantics entailment regime. The approach is based on the computation of lower and upper bounds, but we allow for much more expressive queries than related approaches. In order to optimize the evaluation of possible query answers in the upper but not in the lower bound, we present a query extension approach that uses schema knowledge from the queried ontology to extend the query with additional parts. We show that the resulting query is equivalent to the original one and we use the additional parts that are simple to evaluate for restricting the bounds of subqueries of the initial query. In an empirical evaluation we show that the proposed query extension approach can lead to a significant decrease in the query execution time of up to four orders of magnitude.

PDF Details

ECAI Conference 2014 Conference Paper

Hybrid Query Answering Over OWL Ontologies

Giorgos Stoilos
Giorgos Stamou

Query answering over OWL 2 DL ontologies is an important reasoning task for many modern applications. Unfortunately, due to its high computational complexity, OWL 2 DL systems are still not able to cope with datasets containing billions of data. Consequently, application developers often employ provably scalable systems which only support a fragment of OWL 2 DL and which are, hence, most likely incomplete for the given input. However, this notion of completeness is too coarse since it implies that there exists some query and some dataset for which these systems would miss answers. Nevertheless, there might still be a large number of user queries for which they can compute all the right answers even over OWL 2 DL ontologies. In the current paper, we investigate whether, given a query &Qscr; with only distinguished variables over an OWL 2 DL ontology &Tscr; and a system ans, it is possible to identify in an efficient way if ans is complete for &Qscr; , &Tscr; and every dataset. We give sufficient conditions for (in)completeness and present a hybrid query answering algorithm which uses ans when it is complete, otherwise it falls back to a fully-fledged OWL 2 DL reasoner. However, even in the latter case, our algorithm still exploits ans as much as possible in order to reduce the search space of the OWL 2 DL reasoner. Finally, we have implemented our approach using a concrete system ans and OWL 2 DL reasoner obtaining encouraging results.

Details

ECAI Conference 2006 Conference Paper

General Concept Inclusions inFluzzy Description Logics

Giorgos Stoilos
Umberto Straccia
Giorgos Stamou
Jeff Z. Pan

Fuzzy Description Logics (fuzzy DLs) have been proposed as a language to describe structured knowledge with vague concepts. A major theoretical and computational limitation so far is the inability to deal with General Concept Inclusions (GCIs), which is an important feature of classical DLs. In this paper, we address this issue and develop a calculus for fuzzy DLs with GCIs.

Details