This book provides a broad overview of the key results and frameworks for various NSAI tasks as well as discussing important application areas. This book also covers neuro symbolic reasoning frameworks such as LNN, LTN, and NeurASP and learning frameworks. This would include differential inductive logic programming, constraint learning and deep symbolic policy learning. Additionally, application areas such a visual question answering and natural language processing are discussed as well as topics such as verification of neural networks and symbol grounding. Detailed algorithmic descriptions, example logic programs, and an online supplement that includes instructional videos and slides provide thorough but concise coverage of this important area of AI.
Neuro symbolic artificial intelligence (NSAI) encompasses the combination of deep neural networks with symbolic logic for reasoning and learning tasks. NSAI frameworks are now capable of embedding prior knowledge in deep learning architectures, guiding the learning process with logical constraints, providing symbolic explainability, and using gradient-based approaches to learn logical statements. Several approaches are seeing usage in various application areas.
This book is designed for researchers and advanced-level students trying to understand the current landscape of NSAI research as well as those looking to apply NSAI research in areas such as natural language processing and visual question answering. Practitioners who specialize in employing machine learning and AI systems for operational use will find this book useful as well.
Author(s): Paulo Shakarian; Chitta Baral; Gerardo I. Simari; Bowen Xi; Lahari Pokala
Series: SpringerBriefs in Computer Science
Edition: 1
Publisher: Springer Nature Switzerland
Year: 2023
Language: English
Pages: xii; 119
City: Cham
Tags: Computer Science; Artificial Intelligence; Machine Learning
Foreword
Acknowledgements
Contents
1 New Ideas in Neuro Symbolic Reasoning and Learning
References
2 Brief Introduction to Propositional Logic and Predicate Calculus
2.1 Introduction
2.2 Propositional Logic
2.2.1 Syntax and Semantics
2.2.2 Consistency and Entailment
2.2.3 Entailment and the Fixpoint Operator
2.3 Predicate Calculus
2.4 Conclusion
References
3 Fuzzy and Annotated Logic for Neuro Symbolic Artificial Intelligence
3.1 Introduction
3.2 Generalized Annotated Programs
3.2.1 Syntax of GAPs
3.2.2 Semantics of GAPs
3.2.3 Satisfaction
3.2.4 Theoretical Results for Annotated Logic on a Lower Semi-Lattice
3.3 Fuzzy Logic
3.3.1 Properties of Fuzzy Operators Relevant to NSR
3.3.2 Fuzzy Negation and Considerations for Unit Interval Annotations
3.3.3 T-Norms
3.3.4 T-Conorms
3.3.5 Fuzzy Implication
3.3.6 Aggregators
3.3.7 Product Real Logic
3.3.8 Conorms to Approximate the GAPs Fixpoint Operator
3.4 Chapter Conclusion
References
4 LTN: Logic Tensor Networks
4.1 Introduction and Underlying Language
4.2 From Real Logic to Logic Tensor Networks
4.2.1 Representing Knowledge
4.3 LTN Tasks
4.3.1 Satisfiability and Learning
4.3.2 Querying
4.3.3 Reasoning
4.4 Use Cases
4.4.1 Basic Use Cases: Classification and Clustering
4.4.2 Other Use Cases
4.5 Discussion
4.6 Chapter Conclusions
References
5 Neuro Symbolic Reasoning with Ontological Networks
5.1 Introduction and Underlying Language
5.2 Recursive Reasoning Networks
5.2.1 RRNs: Intuitive Overview
5.2.2 RRNs: A Closer Look
5.3 Use Cases
5.4 Related Approaches
5.5 Chapter Conclusions
References
6 LNN: Logical Neural Networks
6.1 Introduction
6.2 Logic and Inference in LNNs
6.3 Training LNNs
6.4 Discussion
6.5 Chapter Conclusion
References
7 NeurASP
7.1 Introduction
7.2 ASP: Answer Set Programming
7.3 NeurASP
7.3.1 Semantics
7.3.2 Inference in NeurASP
7.3.3 Learning in NeurASP
7.4 Discussion
References
8 Neuro Symbolic Learning with Differentiable Inductive Logic Programming
8.1 Introduction
8.2 ILP Framework
8.2.1 Problem Formulation
8.2.2 Solving ILP Problems
8.3 A Neural Framework for ILP
8.3.1 Loss-Based ILP
8.3.2 Architecture
8.3.3 Complexity
8.3.4 Training Considerations and Empirical Results
8.4 Extensions to δILP
8.5 Conclusion
References
9 Understanding SATNet: Constraint Learning and Symbol Grounding
9.1 Introduction
9.2 SATNet to Learn Instances of MAXSAT
9.2.1 Problem Relaxation
9.2.2 SATNet Forward Pass
9.2.3 Learning the Relaxed Constraint Matrix
9.2.4 Experimental Findings
9.3 Symbol Grounding
9.3.1 Rebuttal on SATNet's Performance on Visual Sudoku
9.4 Discussion
9.5 Chapter Conclusion
References
10 Neuro Symbolic AI for Sequential Decision Making
10.1 Introduction
10.2 Deep Symbolic Policy Learning
10.2.1 Deep Symbolic Regression
10.2.2 Deep Symbolic Policy Learning
10.3 Verifying Neural-Based Models
10.3.1 STLNet
10.4 Discussion
10.5 Chapter Conclusion
References
11 Neuro Symbolic Applications
11.1 Introduction
11.2 Neuro Symbolic Reasoning in Visual Question Answering
11.3 Neuro Symbolic Reasoning involving Natural Language Processing
11.3.1 Concept Learning
11.3.2 Reasoning Over Text
11.3.3 Using ASP in Reasoning Over Text
11.3.4 Solving Logic Grid Puzzles Described in Text
11.4 Neuro-Symbolic Reinforcement Learning
11.5 Chapter Conclusion
References
