Article
Computer Science, Artificial Intelligence
Robin Manhaeve, Sebastijan Dumancic, Angelika Kimmig, Thomas Demeester, Luc De Raedt
Summary: DeepProbLog is a neural probabilistic logic programming language that supports symbolic and subsymbolic representations and inference, program induction, probabilistic programming, and learning from examples. It integrates general-purpose neural networks and expressive probabilistic-logical modeling and reasoning, allowing end-to-end training based on examples.
ARTIFICIAL INTELLIGENCE
(2021)
Review
Computer Science, Artificial Intelligence
Dongran Yu, Bo Yang, Dayou Liu, Hui Wang, Shirui Pan
Summary: In recent years, neural systems have shown effective learning ability and perception intelligence, while symbolic systems exhibit exceptional cognitive intelligence. To combine their advantages, neural-symbolic learning systems are proposed to possess powerful perception and cognition. This paper surveys the advancements in such systems from four perspectives and aims to offer researchers a comprehensive overview and identify promising future directions.
Article
Soil Science
Yu-hao Wu, Hou-biao Li
Summary: In this paper, a new neural symbolic reasoning method called RNNCTPs is proposed, which improves the computational efficiency of traditional symbolic reasoning methods in knowledge graph link prediction by re-filtering the knowledge selection of Conditional Theorem Provers (CTPs), and is less sensitive to the embedding size parameter. RNNCTPs consists of relation selectors and predictors. The relation selectors are trained efficiently and interpretably, allowing the whole model to dynamically generate knowledge for the inference of predictors. In all four datasets, this method shows competitive performance against traditional methods on the link prediction task, and can have higher applicability than CTPs on some datasets.
APPLIED SOIL ECOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Ke Su, Hang Su, Chongxuan Li, Jun Zhu, Bo Zhang
Summary: Neural-symbolic models combine symbolic program execution and deep representation learning for complex visual reasoning tasks. This article proposes a method to incorporate domain knowledge into the learning process of probabilistic neural-symbolic models, which effectively regulates the posterior probability of the structure. Experimental results demonstrate that this method achieves state-of-the-art performance on abstract reasoning datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Hikaru Shindo, Viktor Pfanschilling, Devendra Singh Dhami, Kristian Kersting
Summary: Deep neural learning has shown remarkable performance in learning visual object categorization. However, deep neural networks like CNNs do not explicitly encode objects and their relations, limiting their success on tasks requiring logical understanding of visual scenes. To overcome this, aILP is introduced, a differentiable inductive logic programming framework that represents scenes as logic programs. aILP performs differentiable inductive logic programming on complex visual scenes using gradient descent. Experimental results on Kandinsky patterns and CLEVR-Hans benchmarks demonstrate the accuracy and efficiency of aILP in learning complex visual-logical concepts.
Article
Computer Science, Artificial Intelligence
Kun Gao, Hanpin Wang, Yongzhi Cao, Katsumi Inoue
Summary: The paper introduces a novel differentiable inductive logic programming system called D-LFIT, with characteristics including a small number of parameters, the ability to generate logic programs in a curriculum-learning setting, and linear time complexity for the extraction of trained neural networks.
Article
Computer Science, Software Engineering
Michele Fraccaroli, Evelina Lamma, Fabrizio Riguzzi
Summary: This study improves Bayesian Optimization applied to Deep Neural Networks (DNNs) by analyzing the results on training and validation sets. The resulting system, called Symbolic DNN-Tuner, evaluates and optimizes the network architecture and hyper-parameters using symbolic tuning rules.
Article
Computer Science, Artificial Intelligence
Ashwin Srinivasan, A. Baskar, Tirtharaj Dash, Devanshu Shah
Summary: The key strengths of relational machine learning programs such as Inductive Logic Programming (ILP) lie in their use of expressive logic, domain-specific relations, and human-readable models. However, these methods have not fully utilized the advancements in hardware, software, and algorithms seen in deep neural networks. This paper proposes Compositional Relational Machines (CRMs) as a form of explainable neural network and shows their ability to identify appropriate explanations and provide explanations for black-box models.
Article
Computer Science, Artificial Intelligence
Su He, Xiaofeng Yang, Guosheng Lin
Summary: Visual Grounding is a task that locates a specific object in an image by semantically matching a given linguistic expression. The challenges of this task include mapping linguistic and visual contents and understanding diverse linguistic expressions. In this work, a novel modular network is proposed to match symbolic features and visual features with linguistic information. Additionally, a Residual Attention Parser is designed to address the difficulty of understanding diverse expressions. The model achieves competitive performance on three popular VG datasets.
IMAGE AND VISION COMPUTING
(2022)
Article
Mathematics
Xixi Zhu, Bin Liu, Cheng Zhu, Zhaoyun Ding, Li Yao
Summary: This paper introduces ChunfyReasoner (CFR), an approximate reasoning method that combines neural-symbolic learning with ABox reasoning. By training a neural network model, the CFR speeds up ontology reasoning while ensuring higher reasoning quality.
Article
Computer Science, Artificial Intelligence
Yong-Lu Li, Xinpeng Liu, Xiaoqian Wu, Yizhuo Li, Zuoyu Qiu, Liang Xu, Yue Xu, Hao-Shu Fang, Cewu Lu
Summary: In this article, a novel paradigm is proposed to reformulate the task of human activity understanding. It involves mapping pixels to an intermediate space of atomic activity primitives and then using interpretable logic rules to infer semantics. The proposed framework, Human Activity Knowledge Engine (HAKE), shows superior generalization ability and performance compared to traditional methods.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Theory & Methods
Abdus Salam, Rolf Schwitter, Mehmet A. Orgun
Summary: This survey provides an overview of rule learning systems that can learn the structure of probabilistic rules for uncertain domains, highlighting their usefulness and human-friendly output.
ACM COMPUTING SURVEYS
(2021)
Review
Mathematics
Zefan Zeng, Qing Cheng, Yuehang Si
Summary: The knowledge graph is a powerful form of knowledge representation and management, but biases in our cognitive processes often result in incomplete and erroneous information. Reasoning, particularly logic-based reasoning, is important for supplementing and correcting knowledge graph shortcomings. Despite the growing number of logic-based knowledge graph reasoning methods, there is a lack of systematic classification and analysis. This review examines and classifies these methods into four categories, providing insights into their core concepts, techniques, advantages, and disadvantages, and offering a comparative evaluation of their performance.
Article
Computer Science, Artificial Intelligence
Emile van Krieken, Erman Acar, Frank van Harmelen
Summary: The AI community is increasingly focusing on combining symbolic and neural approaches, with a recent trend towards weakly supervised learning techniques using fuzzy logic operators. The study finds that many logical operators from fuzzy logic literature are unsuitable in differentiable learning settings and introduces a new family of fuzzy implications to address this issue. Empirical results show the possibility of using Differentiable Fuzzy Logics for semi-supervised learning and the need for non-standard combinations of logical operators to achieve performance improvement.
ARTIFICIAL INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Tirtharaj Dash, Ashwin Srinivasan, A. Baskar
Summary: The study introduces a general technique for incorporating multi-relational domain knowledge into Graph Neural Networks using mode-directed inverse entailment (MDIE). This technique, referred to as 'BotGNN', outperforms other methods in terms of computational efficiency and representational versatility, as demonstrated through experiments with various GNN variants and real-world datasets.
Editorial Material
Computer Science, Artificial Intelligence
Haibo He
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2019)
Article
Computer Science, Artificial Intelligence
Son N. Tran, Son Ngo, Artur d'Avila Garcez
NEURAL COMPUTING & APPLICATIONS
(2020)
Article
Computer Science, Artificial Intelligence
Marc van Zee, Dragan Doder, Leendert van der Torre, Mehdi Dastani, Thomas Icard, Eric Pacuit
ARTIFICIAL INTELLIGENCE
(2020)
Article
Computer Science, Artificial Intelligence
Christoph Benzmuller, Xavier Parent, Leendert van der Torre
ARTIFICIAL INTELLIGENCE
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Jinsheng Chen, Beishui Liao, Leendert van der Torre
Summary: Base argumentation is a logic-based instantiation of abstract argumentation, where each base argument is a subset of the given knowledge base. This paper demonstrates that base argumentation meets certain rationality postulates, and is equivalent to deductive argumentation under complete semantics. Due to its simplicity, base argumentation can be viewed as an abstraction of deductive argumentation.
LOGIC AND ARGUMENTATION, CLAR 2021
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Jeremie Dauphin, Tjitze Rienstra, Leendert van der Torre
Summary: Baumann, Brewka, and Ulbricht introduced weak admissibility as an alternative to Dung's notion of admissibility in order to define new semantics for argumentation frameworks. The paper presents six new kinds of semantics based on weak admissibility, providing an initial principle-based analysis of their improvements over existing ones. Such principle-based analysis is useful for selecting semantics for applications, algorithm design, and further research in weak admissibility semantics.
LOGIC AND ARGUMENTATION, CLAR 2021
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Davide Calvaresi, Giovanni Ciatto, Amro Najjar, Reyhan Aydogan, Leon Van Der Torre, Andrea Omicini, Michael Schumacher
Summary: Explainable AI (XAI) has emerged as a set of techniques and methods to interpret machine learning predictors. Current explanation techniques mainly target experts and assume centralised and fully accessible data, which is rarely found in practice. The EXPECTATION project aims to overcome these limitations and construct personalised explanations in decentralised and heterogeneous environments.
EXPLAINABLE AND TRANSPARENT AI AND MULTI-AGENT SYSTEMS, EXTRAAMAS 2021
(2021)
Article
Computer Science, Artificial Intelligence
Llio Humphreys, Guido Boella, Leendert van der Torre, Livio Robaldo, Luigi Di Caro, Sepideh Ghanavati, Robert Muthuri
Summary: This article introduces a semantic role labeling based information extraction system to extract definitions and norms from legislation and represent them as structured norms in legal ontologies. The goal is to make laws more accessible, understandable, and searchable in a legal document management system.
ARTIFICIAL INTELLIGENCE AND LAW
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Liuwen Yu, Reka Markovich, Leendert Van Der Torre
LEGAL KNOWLEDGE AND INFORMATION SYSTEMS
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Beishui Liao, Leendert Van der Torre
COMPUTATIONAL MODELS OF ARGUMENT (COMMA 2020)
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Jeremie Dauphin, Tjitze Rienstra, Leendert Van Der Torre
COMPUTATIONAL MODELS OF ARGUMENT (COMMA 2020)
(2020)
Article
Multidisciplinary Sciences
Christoph Benzmueller, Ali Farjami, David Fuenmayor, Paul Meder, Xavier Parent, Alexander Steen, Leendert van der Torre, Valeria Zahoransky
Article
Computer Science, Artificial Intelligence
Tjitze Rienstra, Chiaki Sakama, Leendert van der Torre, Beishui Liao
ARGUMENT & COMPUTATION
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Beishui Liao, Marija Slavkovik, Leendert van der Torre
AIES '19: PROCEEDINGS OF THE 2019 AAAI/ACM CONFERENCE ON AI, ETHICS, AND SOCIETY
(2019)
Article
Logic
Artur D'Avila Garcez, Marco Gori, Luis C. Lamb, Luciano Serafini, Michael Spranger, Son N. Tran
JOURNAL OF APPLIED LOGICS-IFCOLOG JOURNAL OF LOGICS AND THEIR APPLICATIONS
(2019)
