This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
The Tsetlin Machine Goes Deep: Logical Learning and Reasoning With Graphs
Created by
Haebom
Author
Ole-Christoffer Granmo, Youmna Abdelwahab, Per-Arne Andersen, Paul FA Clarke, Kunal Dumbre, Ylva Gr{\o}nnins{\ae}ter, Vojtech Halenka, Runar Helin, Lei Jiao, Ahmed Khalid, Rebekka Omslandseter, Rupsa Saha, Mayur Shende, Xuan Zhang
Outline
GraphTM is a model that learns interpretable deep clauses from graph-structured inputs. It supports various inputs such as sequential data, lattices, relations, and multi-modality by utilizing graph structures while maintaining the advantages of existing TMs such as pattern recognition and interpretability using simple and flat AND rules, and accuracy similar to deep learning. It recognizes partial graph patterns with exponentially fewer clauses by constructing nested deep clauses through message passing, thereby increasing interpretability and data utilization. It shows higher accuracy or efficiency than existing methods in various fields such as image classification, action co-tracking, recommender systems, and virus genome sequence data, and in particular, it achieved an accuracy that was 3.86% higher than the existing convolutional TM in image classification.
Takeaways, Limitations
•
Takeaways:
◦
We present an interpretable deep learning model for graph-structured data.
◦
Expands applicability to various types of data while maintaining the advantages of existing TM.
◦
Demonstrated excellent performance in various fields such as image classification, action tracking, recommendation systems, and virus genome sequence analysis.
◦
Improved interpretability and data usability through deep segmentation based on message passing.
•
Limitations:
◦
This paper does not provide specific __T10081_____ or future research directions for GraphTM. More detailed experiments and comparative analyses are needed to clarify __T10082_____.
