Authors: Alfie Anthony Treloar, Dany Varghese, Shubhi Verma, Alireza Tamaddoni-Nezhad and Alan Hunter

Abstract: This paper presents preliminary work on integrating symbolic learning and reasoning into autonomous maritime systems using inductive logic programming (ILP). A key challenge in operationalising ILP is bridging the gap between continuous sensing and actuation data and discrete symbolic logic. We propose a framework that enables autonomous vessels to query maritime rules (COLREGs) and learn from human oversight. Using the ILP system PyGol, we demonstrate the learning of COLREG Rule 13 for overtaking situations from discretised bearing data, and further explore the learning of an exception to Rule 15 for crossing situations through examples inspired by case law. These results show the potential for interpretable, legally compliant decision-making and lay the groundwork for learning more complex rules in dynamic maritime environments.

Link: Accepted - The OCEANS 2025 Great Lakes conference

Authors: Shubhi Verma, Dany Varghese,Alfie Anthony Treloar, Alan Hunter and Alireza Tamaddoni-Nezhad

Abstract: Maritime Autonomous Surface Ships (MASS) promise to reduce casualty rates and improve operational efficiency, yet two obstacles impede widespread adoption: the qualitative, often conflicting language of the COLREGs and the opacity of prevailing AI collision-avoidance algorithms. We present a socio-technical decision framework that formalises COLREG hierarchy, including the lex specialis ordering confirmed in Ever Smart v. Alexandra 1, as a tiered rule tree and encodes it in an explainable, auditable knowledge base. Using symbolic logic, the system resolves rule conflicts, logs its reasoning, and outputs a single safe manoeuvre aligned with good seamanship. Three representative scenarios (narrow-channel crossing, cascading multi-vessel conflict, and overtaking in a channel) demonstrate that the framework reproduces expert decisions while exposing a transparent proof trail. The result is a legally coherent foundation for logic-based machine learning using inductive logic programming (ILP) and future maritime autonomous systems trials, advancing the IMO goal of “at least equivalent” safety for unmanned vessels.

Link: Accepted - The OCEANS 2025 Great Lakes conference

Authors: Dany Varghese and Alireza Tamaddoni-Nezhad

Abstract: This paper introduces a symbolic regression framework based on Inductive Logic Programming (ILP) to address the growing demand for interpretable machine learning models in sensitive and regulation-intensive domains. Unlike black-box regressors such as ensemble methods or neural networks, our approach learns human-readable rules that explain how input features relate to output predictions using logic-based representations. We leverage the PyGol, a novel ILP system, to perform multi-class symbolic regression through a one-vs-rest strategy, where continuous targets are either preserved or discretised into symbolic labels. Each label is represented by a distinct set of logic rules defined over feature intervals, facilitating transparent and modular reasoning. A Bayesian-inspired scoring mechanism extends inference to noisy or partially matching instances, enhancing robustness. Through empirical evaluations on benchmark regression datasets, we demonstrate that PyGol achieves competitive predictive performance compared to state-of-the-art regressors while offering superior transparency and traceability. We further present sample learned rules and interpret their behaviour, highlighting the system's explanatory potential. This work affirms the value of ILP-based symbolic models as viable alternatives to black-box approaches, particularly where accountability and decision interpretability are paramount.

Link: Accepted - IJCLR 2025

Authors: Awathy Willson, J Anitha and Dany Varghese

Abstract: Autism Spectrum Disorder (ASD) diagnosis relies on integrating heterogeneous behavioral and cognitive indicators, demanding AI systems that are not only accurate but also interpretable and verifiable. In this study, we present an explainable ASD detection framework based on Inductive Logic Programming (ILP), using phenotypic data from the ABIDE dataset. Unlike black-box models, ILP produces symbolic rules in first-order logic, supporting clinical transparency and auditability. We evaluate ILP against standard machine learning models (e.g., Random Forest, SVM, Gradient Boosting) using 10-fold cross-validation and report competitive accuracy, with ILP demonstrating superior specificity and high precision—critical metrics in clinical screening. We further compare the interpretability of ILP explanations with state-of-the-art post-hoc methods, SHAP and LIME, using a held-out test instance. While all methods identify consistent predictive features, ILP offers globally consistent, human-readable rules that are more accessible to non-expert users. Our findings affirm ILP as a viable and trustworthy alternative for ASD classification, providing both predictive utility and symbolic transparency. Future work will extend this approach to incorporate fMRI-derived features, enabling richer multimodal reasoning in neurodevelopmental diagnostics.

Link: Accepted - IJCLR 2025

Authors: Daniel Cyrus, Dany Varghese, Roman Bauer and Alireza Tamaddoni-Nezhad

Abstract: Learning from small datasets is crucial in biomedical research due to the limited availability of large, annotated data in many domains. Inductive Logic Programming (ILP) offers a robust framework for integrating symbolic reasoning with machine learning, enabling the generation of interpretable models. In this work, we explore the application of numerical symbolic learning approaches to biomedical data using ILP systems such as NumLog, PyGol, and NumSynth. These systems demonstrate superior efficiency in handling numerical features and extracting meaningful rules compared to traditional rule learning and machine learning methods. We evaluate these approaches on two datasets: a neurodegenerative dataset for Alzheimer's disease detection from fundus images and the benchmark Breast Cancer dataset. The results underscore the potential of ILP-based numerical-symbolic learning in identifying complex relationships within biomedical data, providing actionable insights for advancing precision medicine and disease diagnosis.

Link: View Publication

Authors: Dany Varghese and Alireza Tamaddoni-Nezhad

Abstract: Achieving transparency and interpretability in medical artificial intelligence (AI) remains a persistent challenge, particularly as highly accurate but black-box models become more dominant in clinical prediction tasks. A common assumption is that improving model explainability inevitably reduces predictive performance. In this study, we revisit this trade-off by investigating the use of Inductive Logic Programming (ILP) a symbolic machine learning approach-for rule-based medical decision support, and directly compare its predictive performance and interpretability to widely used statistical machine learning models. Using the well-established Cleveland Heart Disease dataset, we systematically evaluate the accuracy and global interpretability of models induced by ILP (implemented via the PyGol system) alongside baseline statistical classifiers. Our results show that, contrary to conventional belief, symbolic models can deliver predictive performance comparable to standard statistical learners, while simultaneously providing concise, human-readable rules that require minimal pre-processing or feature engineering. We illustrate how the learned rules align with known clinical reasoning and offer direct insight into diagnostic decision boundaries. Rather than positioning symbolic learning as a replacement for all black-box methods, we demonstrate its practical potential as a robust, trustworthy, and interpretable alternative for critical medical AI tasks, strengthening the case for transparent decision-making in high-stakes domains.

Link: Accepted - IJCAI 2025 Workshop - Large Language Models and Generative AI for Health Informatics

Authors: Dany Varghese, Daniel Cyrus, Stassa Patsantzis, James Trewern, Alfie Anthony Treloar, Alan Hunter and Alireza Tamaddoni-Nezhad

Abstract: "One-shot learning" traditionally refers to classifying a single instance using a machine learning model pre-trained on extensive datasets. In contrast, Meta Inverse Entailment (MIE), a type of Inductive Logic Programming (ILP), can generate complex logic programs from just a single positive example and minimal background knowledge without prior extensive training. This approach offers a human-centred form of machine learning that is more controllable, reliable, and comprehensible due to its small training data size and the inherent interpretability of logic programs. We use PyGol, a Python-based implementation of Meta Inverse Entailment, and compare its performance with ExpGen-PPO, a leading deep reinforcement learning system. Our experiments focus on two domains: maze-solving and obstacle avoidance for mobile robotics. In both domains, we first train the systems in simplified environments without obstacles and then test their ability to generalise to more complex environments with obstacles. Our results show that PyGol effectively learns generalisable solutions from a single example in both domains, whereas ExpGen-PPO requires more training and significantly more exploration to achieve similar performance.

Link: In-Press

Authors: Dany Varghese, Ghazal Afroozi Milani, and Alireza Tamaddoni-Nezhad

Abstract: Large Language Models (LLMs) have achieved remarkable success in various natural language processing tasks. However, their ability to perform tasks requiring formal representation and reasoning remains limited. This paper explores the integration of Meta Inverse Entailment (MIE) with LLMs to enhance their reasoning capabilities. In our experiments, we examine a hybrid GPT-MIE model on a simplified natural language grammar. The results suggest that the accuracy of GPT is significantly improved when it incorporates the grammar learned using MIE. This hybrid approach demonstrates the potential of combining LLMs' linguistic proficiency with MIE's rigorous formalism, leading to better performance in tasks demanding logical representation and reasoning.

Link: In-Press

Authors: Daniel Cyrus, Dany Varghese and Alireza Tamaddoni-Nezhad

Abstract: In this paper, we present NumLog, an Inductive Logic Programming (ILP) system designed for feature range discovery. NumLog generates quantitative rules with clear confidence bounds to discover feature-range values from examples. Our approach focuses on generating rules with minimal complexity from numerical values, ensuring the assessment of methods that could impact accuracy and comprehensibility. Traditional ILP systems, especially those intersecting with computer vision, struggle with numerical data. This convergence presents unique challenges, often hindering the generation of meaningful insights due to the limited capabilities of conventional ILP systems to handle numerical values. NumLog stands out by incorporating an advanced range discovery mechanism that generates low-complexity rules while maintaining high accuracy and comprehensibility. This enhancement significantly improves interpretability, promoting more effective human-machine learning collaboration. We compare NumLog with the state-of-the-art ILP systems such as NumSynth and Aleph and conduct comprehensive experiments on several datasets. We evaluated our approach by measuring accuracy, precision, F1 score, and rule complexity to demonstrate the effectiveness of the methodology.

Link: View Publication

Authors: Dany Varghese, Roman Bauer and Alireza Tamaddoni-Nezhad

Abstract: Traditional machine learning methods heavily rely on large amounts of labelled data for effective generalisation, posing a challenge in few-shot learning scenarios. In many real-world applications, acquiring large amounts of training data can be difficult or impossible. This paper presents an efficient and explainable method for few-shot learning from images using inductive logic programming (ILP). ILP utilises logical representations and reasoning to capture complex relationships and generalise from sparse data. We demonstrate the effectiveness of our proposed ILP-based approach through an experimental evaluation focused on detecting neurodegenerative diseases from fundus images. By extending our previous work on neurodegenerative disease detection, including Alzheimer’s disease, Parkinson’s disease, and vascular dementia disease, we achieve improved explainability in identifying these diseases using fundus images collected from the UK Biobank dataset. The logical representation and reasoning inherent in ILP enhances the interpretability of the detection process. The results highlight the efficacy of ILP in few-shot learning scenarios, showcasing its remarkable generalisation performance compared to a range of other machine learning algorithms. This research contributes to the field of few-shot learning using ILP and paves the way for addressing challenging real-world problems.

Link: View Publication

Authors: Didac Barroso-Bergada, Alireza Tamaddoni-Nezhad, Dany Varghese, Corinne Vacher, Nika Galic, Valérie Laval, Frédéric Suffert and David A Bohan

Abstract: The functional diversity of microbial communities emerges from a combination of the great number of species and the many interaction types, such as competition, mutualism, predation or parasitism, in microbial ecological networks. Understanding the relationship between microbial networks and the functions delivered by the microbial communities is a key challenge for microbial ecology, particularly as so many of these interactions are difficult to observe and characterise. We believe that this ’Dark Web’ of interactions could be unravelled using an explainable machine learning approach, called Abductive/Inductive Logic Programming (A/ILP) in the R package InfIntE, which uses mechanistic rules (interaction hypotheses) to infer directly the network structure and interaction types. Here we attempt to unravel the dark web of the plant microbiome in metabarcoding data sampled from the grapevine foliar microbiome. Using synthetic, simulated data, we first show that it is possible to satisfactorily reconstruct microbial networks using explainable machine learning. Then we confirm that the dark web of the grapevine microbiome is diverse, being composed of a range of interaction types consistent with the literature. This first attempt to use explainable machine learning to infer microbial interaction networks advances our understanding of the ecological processes that occur in microbial communities and allows us to hypothesise specific types of interaction within the grapevine microbiome. This work will have potentially valuable applications, such as the discovery of antagonistic interactions that might be used to identify potential biological control agents within the microbiome.

Link: View Publication

Authors: Dany Varghese, Uzma Patel, Paul Krause and Alireza Tamaddoni-Nezhad

Abstract: Plant diseases are one of the main causes of crop loss in agriculture. Machine Learning, in particular statistical and neural nets (NNs) approaches, have been used to help farmers identify plant diseases. However, since new diseases continue to appear in agriculture due to climate change and other factors, we need more data-efficient approaches to identify and classify new diseases as early as possible. Even though statistical machine learning approaches and neural nets have demonstrated state-of-the-art results on many classification tasks, they usually require a large amount of training data. This may not be available for emergent plant diseases. So, data-efficient approaches are essential for an early and precise diagnosis of new plant diseases and necessary to prevent the disease’s spread. This study explores a data-efficient Inductive Logic Programming (ILP) approach for plant disease classification. We compare some ILP algorithms (including our new implementation, PyGol) with several statistical and neural-net based machine learning algorithms on the task of tomato plant disease classification with varying sizes of training data set (6, 10, 50 and 100 training images per disease class). The results suggest that ILP outperforms other learning algorithms and this is more evident when fewer training data are available.

Link: View Publication

Authors: Dany Varghese, Didac Barroso-Bergada, David A Bohan and Alireza Tamaddoni-Nezhad

Abstract: Abductive reasoning plays an essential part in day-to-day problem-solving. It has been considered a powerful mechanism for hypothetical reasoning in the presence of incomplete knowledge; a form of “common sense” reasoning. In machine learning, abduction is viewed as a conceptual method in which data and the bond that jointly brings the different types of inference. The traditional Mode-Directed Inverse Entailment (MDIE) based systems such as Progol and Aleph for the abduction were not data-efficient since their execution time with the large dataset was too long. We present a new abductive learning procedure using Meta Inverse Entailment (MIE). MIE is similar to Mode-Directed Inverse Entailment (MDIE) but does not require user-defined mode declarations. In this paper, we use an implementation of MIE in Python called PyGol. We evaluate and compare this approach to reveal the microbial interactions in the ecosystem with state-of-art-of methods for abduction, such as Progol and Aleph. Our results show that PyGol has comparable predictive accuracies but is significantly faster than Progol and Aleph.

Link: View Publication

Authors: Dany Varghese,Roman Bauer, Daniel Baxter-Beard, Stephen Muggleton and Alireza Tamaddoni-Nezhad

Abstract: Unlike most computer vision approaches, which depend on hundreds or thousands of training images, humans can typically learn from a single visual example. Humans achieve this ability using background knowledge. Rule-based machine learning approaches such as Inductive Logic Programming (ILP) provide a framework for incorporating domain specific background knowledge. These approaches have the potential for human-like learning from small data or even one-shot learning, i.e. learning from a single positive example. By contrast, statistics based computer vision algorithms, including Deep Learning, have no general mechanisms for incorporating background knowledge. This paper presents an approach for one-shot rule learning called One-Shot Hypothesis Derivation (OSHD) based on using a logic program declarative bias. We apply this approach to two challenging human-like computer vision tasks: 1) Malayalam character recognition and 2) neurological diagnosis using retinal images. We compare our results with a state-of-the-art Deep Learning approach, called Siamese Network, developed for one-shot learning. The results suggest that our approach can generate human-understandable rules and outperforms the deep learning approach with a significantly higher average predictive accuracy.

Link: View Publication

Authors: Dany Varghese and Alireza Tamaddoni-Nezhad

Abstract: Unlike most of computer vision approaches which depend on hundreds or thousands of training images, humans can typically learn from a single visual example. Humans achieve this ability using background knowledge. Rule-based machine learning approaches such as In-ductive Logic Programming (ILP) provide a framework for incorporating domain specific background knowledge. These approaches have the potential for human-like learning from small data or even one-shot learning, i.e. learning from a single positive example. By contrast, statistics based computer vision algorithms, including Deep Learning, have no general mechanisms for incorporating background knowledge. In this paper, we present an approach for one-shot rule learning called One-Shot Hypothesis Derivation (OSHD) which is based on using a logic program declarative bias. We apply this approach to the challenging task of Malayalam character recognition. This is a challenging task due to spherical and complex structure of Malayalam handwritten language. Unlike for other languages, there is currently no efficient algorithm for Malayalam handwritten recognition. We compare our results with a state-of-the-art Deep Learning approach, called Siamese Network, which has been developed for one-shot learning. The results suggest that our approach can generate human-understandable rules and also outperforms the deep learning approach with a significantly higher average predictive accuracy.

Link: View Publication