Publications
243 results found
Cunnington D, Law M, Lobo J, et al., 2023, FFNSL: feed-forward neural-symbolic learner, Machine Learning, Vol: 112, Pages: 515-569, ISSN: 0885-6125
Logic-based machine learning aims to learn general, interpretable knowledge in a data-efficient manner. However, labelled data must be specified in a structured logical form. To address this limitation, we propose a neural-symbolic learning framework, called Feed-Forward Neural-Symbolic Learner (FFNSL), that integrates a logic-based machine learning system capable of learning from noisy examples, with neural networks, in order to learn interpretable knowledge from labelled unstructured data. We demonstrate the generality of FFNSL on four neural-symbolic classification problems, where different pre-trained neural network models and logic-based machine learning systems are integrated to learn interpretable knowledge from sequences of images. We evaluate the robustness of our framework by using images subject to distributional shifts, for which the pre-trained neural networks may predict incorrectly and with high confidence. We analyse the impact that these shifts have on the accuracy of the learned knowledge and run-time performance, comparing FFNSL to tree-based and pure neural approaches. Our experimental results show that FFNSL outperforms the baselines by learning more accurate and interpretable knowledge with fewer examples.
Russo A, Dickens L, Stromfelt H, et al., 2022, Formalizing Coherence and Consistency Applied to Transfer Learning in Neuro-Symbolic Autoencoders, Thirty-sixth Conference on Neural Information Processing Systems
Aspis Y, Broda K, Lobo J, et al., 2022, Embed2Sym - scalable neuro-symbolic reasoning via clustered embeddings, The 19th International Conference on Principles of Knowledge Representation and Reasoning, Publisher: K Proceedings, Pages: 421-431
Neuro-symbolic reasoning approaches proposed in recent years combine a neural perception component with a symbolic reasoning component to solve a downstream task. By doing so, these approaches can provide neural networks with symbolic reasoning capabilities, improve their interpretability and enable generalization beyond the training task. However, this often comes at the cost of poor training time, with potential scalability issues. In this paper, we propose a scalable neuro-symbolic approach, called Embed2Sym. We complement a two-stage (perception and reasoning) neural network architecture designed to solve a downstream task end-to-end with a symbolic optimisation method for extracting learned latent concepts. Specifically, the trained perception network generates clusters in embedding space that are identified and labelled using symbolic knowledge and a symbolic solver. With the latent concepts identified, a neuro-symbolic model is constructed by combining the perception network with the symbolic knowledge of the downstream task, resulting in a model that is interpretable and transferable. Our evaluation shows that Embed2Sym outperforms state-of-the-art neuro-symbolic systems on benchmark tasks in terms of training time by several orders of magnitude while providing similar if not better accuracy.
Mitchener L, Tuckey D, Crosby M, et al., 2022, Detect, understand, act: a neuro-symbolic hierarchical reinforcement learning framework (extended abstract), THE 31ST INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, Publisher: IJCAI, Pages: 5314-5318, ISSN: 1045-0823
We introduce Detect, Understand, Act (DUA), a neuro-symbolic reinforcement learning framework. The Detect component is composed of a traditional computer vision object detector and tracker. The Act component houses a set of options, high-level actions enacted by pre-trained deep reinforcement learning (DRL) policies. The Understand component provides a novel answer set programming (ASP) paradigm for effectively learning symbolic meta-policies over options using inductive logic programming (ILP). We evaluate our framework on the Animal-AI (AAI) competition testbed, a set of physical cognitive reasoning problems. Given a set of pre-trained DRL policies, DUA requires only a few examples to learn a meta-policy that allows it to improve the state-of-the-art on multiple of the most challenging categories from the testbed. DUA constitutes the first holistic hybrid integration of computer vision, ILP and DRL applied to an AAI-like environment and sets the foundations for further use of ILP in complex DRL challenges.
Law M, Broda K, Russo A, 2022, Search space expansion for efficient incremental inductive logic programming from streamed data, THE 31ST INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, Pages: 2697-2704, ISSN: 1045-0823
In the past decade, several systems for learning Answer Set Programs (ASP) have been proposed, including the recent FastLAS system. Compared to other state-of-the-art approaches to learning ASP, FastLAS is more scalable, as rather than computing the hypothesis space in full, it computes a much smaller subset relative to a given set of examples that is nonetheless guaranteed to contain an optimal solution to the task (called an OPT-sufficient subset). On the other hand, like many other Inductive Logic Programming (ILP) systems, FastLAS is designed to be run on a fixed learning task meaning that if new examples are discovered after learning, the whole process must be run again. In many real applications, data arrives in a stream. Rerunning an ILP system from scratch each time new examples arrive is inefficient. In this paper we address this problem by presenting IncrementalLAS, a system that uses a new technique, called hypothesis space expansion, to enable a FastLAS-like OPT-sufficient subset to be expanded each time new examples are discovered. We prove that this preserves FastLAS's guarantee of finding an optimal solution to the full task (including the new examples), while removing the need to repeat previous computations. Through our evaluation, we demonstrate that running IncrementalLAS on tasks updated with sequences of new examples is significantly faster than re-running FastLAS from scratch on each updated task.
Mitchener L, Tuckey D, Crosby M, et al., 2022, Detect, understand, act: a neuro-symbolic hierarchical reinforcement learning framework, Machine Learning, Vol: 111, Pages: 1523-1549, ISSN: 0885-6125
In this paper we introduce Detect, Understand, Act (DUA), a neuro-symbolic reinforcement learning framework. The Detect component is composed of a traditional computer vision object detector and tracker. The Act component houses a set of options, high-level actions enacted by pre-trained deep reinforcement learning (DRL) policies. The Understand component provides a novel answer set programming (ASP) paradigm for symbolically implementing a meta-policy over options and effectively learning it using inductive logic programming (ILP). We evaluate our framework on the Animal-AI (AAI) competition testbed, a set of physical cognitive reasoning problems. Given a set of pre-trained DRL policies, DUA requires only a few examples to learn a meta-policy that allows it to improve the state-of-the-art on multiple of the most challenging categories from the testbed. DUA constitutes the first holistic hybrid integration of computer vision, ILP and DRL applied to an AAI-like environment and sets the foundations for further use of ILP in complex DRL challenges.
Tuckey D, Broda K, Russo A, 2022, A semantics for probabilistic answer set programs with incomplete stochastic knowledge, CEUR Workshop, Publisher: CEUR Workshop Proceedings, Pages: 1-14, ISSN: 1613-0073
Some probabilistic answer set programs (PASP) semantics assign probabilities to sets of answer sets and implicitly assume these answer sets to be equiprobable. While this is a common choice in probability theory, it leads to unnatural behaviours with PASPs. We argue that the user should have a level of control over what assumption is used to obtain a probability distribution when the stochastic knowledge is incomplete. To this end, we introduce the Incomplete Knowledge Semantics (IKS) for probabilistic answer set programs. We take inspiration from the field of decision making under ignorance. Given a cost function, represented by a user-defined ordering over answer sets through weak constraints, we use the notion of Ordered Weighted Averaging (OWA) operator to distribute the probability over a set of answer sets accordingly to the user’s level of optimism. The more optimistic (or pessimistic) a user is, the more (or less) probability is assigned to the more optimal answer sets. We present an implementation and showcase the behaviour of this semantics on simple examples. We also highlight the impact that different OWA operators have on weight learning, showing that the equiprobability assumption is not always the best option.
Russo A, Law M, Cunnington D, et al., 2022, Logic-Based Machine Learning: Recent Advances and Their Role in Neuro-Symbolic AI, 16th International Conference on Logic Programming and Non-Monotonic Reasoning (LPNMR), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: XVIII-XXI, ISSN: 0302-9743
Law M, Russo A, Broda K, et al., 2021, Scalable non-observational predicate learning in ASP, IJCAI, Publisher: IJCAI, Pages: 1936-1943, ISSN: 1045-0823
Recently, novel ILP systems under the answer set semantics have been proposed, some of which are robust to noise and scalable over large hypothesis spaces. One such system is FastLAS, which is significantly faster than other state-of-the-art ASP-based ILP systems. FastLAS is, however, only capable of Observational Predicate Learning (OPL),where the learned hypothesis defines predicates that are directly observed in the examples. It cannot learn knowledge that is indirectly observable, such as learning causes of observed events. This class of problems, known as non-OPL, is known to be difficult to handle in the context of non-monotonic semantics. Solving non-OPL learning tasks whilst preserving scalability is a challenging open problem. We address this problem with a new abductive method for translating examples of a non-OPL task to a set of examples, called possibilities, such that the original example is covered if at least one of the possibilities is covered. This new method al-lows an ILP system capable of performing OPL tasks to be “upgraded” to solve non-OPL tasks. In particular, we present our new FastNonOPL system, which upgrades FastLAS with the new possibility generation. We compare it to other state-of-the-art ASP-based ILP systems capable of solving non-OPL tasks, showing that FastNonOPL is significantly faster, and in many cases more accurate, than these other systems.
Koschate M, Naserian E, Dickens L, et al., 2021, ASIA: Automated Social Identity Assessment using linguistic style, Behavior Research Methods, Vol: 53, Pages: 1762-1781, ISSN: 1554-351X
The various group and category memberships that we hold are at the heart of who we are. They have been shown to affect our thoughts, emotions, behavior, and social relations in a variety of social contexts, and have more recently been linked to our mental and physical well-being. Questions remain, however, over the dynamics between different group memberships and the ways in which we cognitively and emotionally acquire these. In particular, current assessment methods are missing that can be applied to naturally occurring data, such as online interactions, to better understand the dynamics and impact of group memberships in naturalistic settings. To provide researchers with a method for assessing specific group memberships of interest, we have developed ASIA (Automated Social Identity Assessment), an analytical protocol that uses linguistic style indicators in text to infer which group membership is salient in a given moment, accompanied by an in-depth open-source Jupyter Notebook tutorial (https://github.com/Identity-lab/Tutorial-on-salient-social-Identity-detection-model). Here, we first discuss the challenges in the study of salient group memberships, and how ASIA can address some of these. We then demonstrate how our analytical protocol can be used to create a method for assessing which of two specific group memberships—parents and feminists—is salient using online forum data, and how the quality (validity) of the measurement and its interpretation can be tested using two further corpora as well as an experimental study. We conclude by discussing future developments in the field.
Furelos-Blanco D, Law M, Jonsson A, et al., 2021, Induction and exploitation of subgoal automata for reinforcement learning, Journal of Artificial Intelligence Research, Vol: 70, Pages: 1031-1116, ISSN: 1076-9757
In this paper we present ISA, an approach for learning and exploiting subgoals in episodic reinforcement learning (RL) tasks. ISA interleaves reinforcement learning with the induction of a subgoal automaton, an automaton whose edges are labeled by the task’s subgoals expressed as propositional logic formulas over a set of high-level events. A subgoal automaton also consists of two special states: a state indicating the successful completion of the task, and a state indicating that the task has finished without succeeding. A state-of-the-art inductive logic programming system is used to learn a subgoal automaton that covers the traces of high-level events observed by the RL agent. When the currently exploited automaton does not correctly recognize a trace, the automaton learner induces a new automaton that covers that trace. The interleaving process guarantees the induction of automata with the minimum number of states, and applies a symmetry breaking mechanism to shrink the search space whilst remaining complete. We evaluate ISA in several gridworld and continuous state space problems using different RL algorithms that leverage the automaton structures. We provide an in-depth empirical analysis of the automaton learning performance in terms of the traces, the symmetry breaking and specific restrictions imposed on the final learnable automaton. For each class of RL problem, we show that the learned automata can be successfully exploited to learn policies that reach the goal, achieving an average reward comparable to the case where automata are not learned but handcrafted and given beforehand.
Cingillioglu N, Russo A, 2021, pix2rule: End-to-end Neuro-symbolic Rule Learning, 15th International Workshop on Neural-Symbolic Learning and Reasoning (NeSy) as part of the 1st International Joint Conference on Learning and Reasoning (IJCLR), Publisher: RWTH AACHEN, Pages: 15-56, ISSN: 1613-0073
Cunnington D, Law M, Russo A, et al., 2021, Towards Neural-Symbolic Learning to support Human-Agent Operations, 24th IEEE International Conference on Information Fusion (FUSION), Publisher: IEEE, Pages: 223-230
Furelos-Blanco D, Law M, Jonsson A, et al., 2021, Induction and Exploitation of Subgoal Automata for Reinforcement Learning, JOURNAL OF ARTIFICIAL INTELLIGENCE RESEARCH, Vol: 70, Pages: 1031-1116, ISSN: 1076-9757
- Author Web Link
- Cite
- Citations: 4
Al-Negheimish H, Madhyastha P, Russo A, 2021, Numerical reasoning in machine reading comprehension tasks: are we there yet?, Conference on Empirical Methods in Natural Language Processing (EMNLP), Publisher: ASSOC COMPUTATIONAL LINGUISTICS-ACL, Pages: 9643-9649
Drozdov A, Law M, Lobo J, et al., 2021, Online Symbolic Learning of Policies for Explainable Security, 3rd EEE International Conference on Trust, Privacy and Security in Intelligent Systems and Applications (TPS-ISA), Publisher: IEEE COMPUTER SOC, Pages: 269-278
- Author Web Link
- Cite
- Citations: 1
Preece A, Braines D, Cerutti F, et al., 2021, Coalition Situational Understanding via Explainable Neuro-Symbolic Reasoning and Learning, Conference on Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications III, Publisher: SPIE-INT SOC OPTICAL ENGINEERING, ISSN: 0277-786X
Al-Negheimish H, Madhyastha P, Russo A, 2021, Discrete Reasoning Templates for Natural Language Understanding, 16th Conference of the European-Chapter-of-the-Association-for-Computational-Linguistics (EACL), Publisher: ASSOC COMPUTATIONAL LINGUISTICS-ACL, Pages: 80-87
Stromfelt H, Dickens L, Garcez AD, et al., 2021, Coherent and Consistent Relational Transfer Learning with Auto-encoders, 15th International Workshop on Neural-Symbolic Learning and Reasoning (NeSy) as part of the 1st International Joint Conference on Learning and Reasoning (IJCLR), Publisher: RWTH AACHEN, Pages: 176-192, ISSN: 1613-0073
Cingillioglu N, Russo A, 2020, Learning invariants through soft unification, Advances in Neural Information Processing Systems 33, ISSN: 1049-5258
Human reasoning involves recognising common underlying principles across many examples. The by-products of such reasoning are invariants that capture patterns such as “if someone went somewhere then they are there”, expressed using variables “someone” and “somewhere” instead of mentioning specific people or places. Humans learn what variables are and how to use them at a young age. This paper explores whether machines can also learn and use variables solely from examples without requiring human pre-engineering. We propose Unification Networks, an end-to-end differentiable neural network approach capable of lifting examples into invariants and using those invariants to solve a given task. The core characteristic of our architecture is soft unification between examples that enables the network to generalise parts of the input into variables, thereby learning invariants. We evaluate our approach on five datasets to demonstrate that learning invariants captures patterns in the data and can improve performance over baselines.
Ricca F, Russo A, 2020, Introduction to the 36th international conference on logic programming special issue II, THEORY AND PRACTICE OF LOGIC PROGRAMMING, Vol: 20, Pages: 815-817, ISSN: 1471-0684
This is the second issue with the selected papers of the 36th International Conference on LogicProgramming (ICLP 2020), held virtually in Rende (CS), Italy, from September 20 to September25, 2020. The two issues contain 27 papers selected from several tracks of the conference program for publication in Theory and Practice of Logic Programming. The preceding issue of thisvolume of the journal contains a detailed editorial by the conference chairs (?), as well as thefollowing fourteen papers selected for publication.
Abu Jabal A, Bertino E, Lobo J, et al., 2020, Polisma - a framework for learning attribute-based access control policies, European Symposium on Research in Computer Security, Publisher: Springer International Publishing, Pages: 523-544, ISSN: 0302-9743
Attribute-based access control (ABAC) is being widely adopted due to its flexibility and universality in capturing authorizations in terms of the properties (attributes) of users and resources. However, specifying ABAC policies is a complex task due to the variety of such attributes. Moreover, migrating an access control system adopting a low-level model to ABAC can be challenging. An approach for generating ABAC policies is to learn them from data, namely from logs of historical access requests and their corresponding decisions. This paper proposes a novel framework for learning ABAC policies from data. The framework, referred to as Polisma, combines data mining, statistical, and machine learning techniques, capitalizing on potential context information obtained from external sources (e.g., LDAP directories) to enhance the learning process. The approach is evaluated empirically using two datasets (real and synthetic). Experimental results show that Polisma is able to generate ABAC policies that accurately control access requests and outperforms existing approaches.
Aspis Y, Broda K, Russo A, et al., 2020, Stable and supported semantics in continuous vector spaces, 17th International Conference on Principles of Knowledge Representation and Reasoning, KR 2020, Publisher: https://proceedings.kr.org/2020/7/, Pages: 58-67
We introduce a novel approach for the computation of stable and supported models of normal logic programs in continuous vector spaces by a gradient-based search method. Specifically, the application of the immediate consequence operator of a program reduct can be computed in a vector space. To do this, Herbrand interpretations of a propositional program are embedded as 0-1 vectors in R and program reducts are represented as matrices in ℝ . Using these representations we prove that the underlying semantics of a normal logic program is captured through matrix multiplication and a differentiable operation. As supported and stable models of a normal logic program can now be seen as fixed points in a continuous space, non-monotonic deduction can be performed using an optimisation process such as Newton's method. We report the results of several experiments using synthetically generated programs that demonstrate the feasibility of the approach and highlight how different parameter values can affect the behaviour of the system. N N×N
Ricca F, Russo A, 2020, Introduction to the 36th International Conference on Logic Programming Special Issue I, THEORY AND PRACTICE OF LOGIC PROGRAMMING, Vol: 20, Pages: 587-592, ISSN: 1471-0684
- Author Web Link
- Cite
- Citations: 1
Casale G, Artac M, van den Heuvel W-J, et al., 2020, RADON: Rational decomposition and orchestration for serverless computing, SICS Software-Intensive Cyber-Physical Systems, Vol: 35, Pages: 77-87, ISSN: 2524-8529
Emerging serverless computing technologies, such as function as a service (FaaS), enable developers to virtualize the internal logic of an application, simplifying the management of cloud-native services and allowing cost savings through billing and scaling at the level of individual functions. Serverless computing is therefore rapidly shifting the attention of software vendors to the challenge of developing cloud applications deployable on FaaS platforms.In this vision paper, we present the research agendaof the RADON project (http://radon-h2020.eu),which aims to develop a model-driven DevOps framework for creating and managing applications based onserverless computing. RADON applications will consist of fine-grained and independent microservices that canefficiently and optimally exploit FaaS and containertechnologies. Our methodology strives to tackle complexity in designing such applications, including the solution of optimal decomposition, the reuse of serverlessfunctions as well as the abstraction and actuation ofevent processing chains, while avoiding cloud vendorlock-in through models.
Gomoluch P, Alrajeh D, Russo A, et al., 2020, Learning neural search policies for classical planning, International Conference on Automated Planning and Scheduling (ICAPS) 2020, Publisher: AAAI, Pages: 522-530, ISSN: 2334-0835
Heuristic forward search is currently the dominant paradigmin classical planning. Forward search algorithms typicallyrely on a single, relatively simple variation of best-first searchand remain fixed throughout the process of solving a plan-ning problem. Existing work combining multiple search tech-niques usually aims at supporting best-first search with anadditional exploratory mechanism, triggered using a hand-crafted criterion. A notable exception is very recent workwhich combines various search techniques using a trainablepolicy. That approach, however, is confined to a discrete ac-tion space comprising several fixed subroutines.In this paper, we introduce a parametrized search algorithmtemplate which combines various search techniques withina single routine. The template’s parameter space defines aninfinite space of search algorithms, including, among others,BFS, local and random search. We then propose a neural ar-chitecture for designating the values of the search parametersgiven the state of the search. This enables expressing neuralsearch policies that change the values of the parameters asthe search progresses. The policies can be learned automat-ically, with the objective of maximizing the planner’s per-formance on a given distribution of planning problems. Weconsider a training setting based on a stochastic optimizationalgorithm known as thecross-entropy method(CEM). Exper-imental evaluation of our approach shows that it is capable offinding effective distribution-specific search policies, outper-forming the relevant baselines.
Law M, Russo A, Bertino E, et al., 2020, FastLAS: scalable inductive logic programming incorporating domain-specific optimisation criteria, The Thirty-Fourth AAAI Conference on Artificial Intelligence (AAAI-2020), Publisher: Association for the Advancement of ArtificialIntelligence, Pages: 2877-2885
Inductive Logic Programming (ILP) systems aim to find a setof logical rules, called a hypothesis, that explain a set of ex-amples. In cases where many such hypotheses exist, ILP sys-tems often bias towards shorter solutions, leading to highlygeneral rules being learned. In some application domains likesecurity and access control policies, this bias may not be de-sirable, as when data is sparse more specific rules that guaran-tee tighter security should be preferred. This paper presents anew general notion of ascoring functionover hypotheses thatallows a user to express domain-specific optimisation criteria.This is incorporated into a new ILP system, calledFastLAS,that takes as input a learning task and a customised scoringfunction, and computes an optimal solution with respect tothe given scoring function. We evaluate the accuracy of Fast-LAS over real-world datasets for access control policies andshow that varying the scoring function allows a user to tar-get domain-specific performance metrics. We also compareFastLAS to state-of-the-art ILP systems, using the standardILP bias for shorter solutions, and demonstrate that FastLASis significantly faster and more scalable.
Verma DC, Bertino E, Russo A, et al., 2020, Policy Based Ensembles for Multi Domain Operations, Conference on Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications II Part of SPIE Defense + Commercial Sensing Conference, Publisher: SPIE-INT SOC OPTICAL ENGINEERING, ISSN: 0277-786X
- Author Web Link
- Cite
- Citations: 2
Furelos-Blanco D, Law M, Jonsson A, et al., 2020, Induction of Subgoal Automata for Reinforcement Learning, 34th AAAI Conference on Artificial Intelligence / 32nd Innovative Applications of Artificial Intelligence Conference / 10th AAAI Symposium on Educational Advances in Artificial Intelligence, Publisher: ASSOC ADVANCEMENT ARTIFICIAL INTELLIGENCE, Pages: 3890-3897, ISSN: 2159-5399
- Author Web Link
- Cite
- Citations: 7
Tuckey D, Broda K, Russo A, 2020, Towards Structure Learning under the Credal Semantics, ISSN: 1613-0073
We present the Credal-FOIL system for structure learning of probabilistic logic programs under the credal semantics. The credal semantics is a generalisation of the distribution semantics based on the answer set semantics. Our learning approach takes a set of examples that are atoms with target lower and upper bounds probabilities and a background knowledge that can have negative loops. We define accuracy in this setting and learn a set of normal rules without loops that maximises this notion of accuracy. We showcase the system on two proof-of-concept examples.
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.