Hyperbolic Learning for Structured Data, Knowledge, and Memory: A Tutorial will take place at KDD 2026. The tutorial focuses on how hyperbolic geometry can support structured data organization, retrieval, knowledge interfaces, and memory layers in modern foundation-model systems.

Important Information

  • Tutorial title: Hyperbolic Learning for Structured Data, Knowledge, and Memory: A Tutorial
  • Date: Sunday, August 9, 2026
  • Time: 8:00 AM - 12:00 PM
  • Conference: KDD 2026
  • Venue: Jeju, Korea
  • Format: Half-day lecture-style tutorial with interactive questions and discussion

Tutorial Introduction

Foundation models are increasingly deployed as data-and-memory systems that combine pretrained parameters with retrieval interfaces, external knowledge stores, persistent memory, and continually updated corpora. Many KDD problems, including recommendation, search, temporal modeling, knowledge discovery, enterprise knowledge systems, and AI for science, involve long-tail, hierarchical, and relational data that are not naturally matched to purely Euclidean latent spaces.

This tutorial introduces hyperbolic learning as a practical geometric framework for structured data, knowledge, and memory layers. The focus is on when curved geometry improves data organization, retrieval interfaces, memory updates, adaptation, and deployment behavior, and how these gains should be evaluated in KDD systems.

Tutorial Topics

The tutorial will cover:

  • Hyperbolic geometry essentials, including Poincare ball and hyperboloid models, geodesic distance, exponential and logarithmic maps, and optimization in curved spaces
  • Hyperbolic data layers for representation, indexing, retrieval, attention, normalization, and numerical stability
  • Hyperbolic memory and agent systems, including retrieval-aware LLMs, non-parametric memory, agent memory management, model editing, and unlearning
  • Multimodal, generative, and scientific data systems, including vision-language modeling, hierarchical multimodal understanding, and AI-for-science scenarios
  • KDD-native applications and evaluation in recommendation, temporal and relational learning, enterprise knowledge systems, scaling strategies, and responsible deployment

Target Audience and Prerequisites

The tutorial targets researchers, practitioners, and graduate students interested in foundation models, hyperbolic learning, geometric deep learning, and KDD applications involving structured data, knowledge, retrieval, or memory. No differential geometry background is required. Familiarity with basic machine learning, deep learning, and standard foundation-model architectures is sufficient.

Tutors

All tutors are planned to participate in person for this tutorial.

  • Jiahong Liu (The Chinese University of Hong Kong): Ph.D. candidate whose research focuses on hyperbolic geometry for foundation models, recommender systems, and federated learning.
  • Menglin Yang (HKUST(GZ)): Assistant Professor in AI Thrust at HKUST (Guangzhou). His research interests include hyperbolic geometric learning, graph representation learning, and foundation models.
  • Irwin King (The Chinese University of Hong Kong): Pro-Vice-Chancellor (Education) and Professor whose research spans geometric machine learning, hyperbolic representation learning, data mining, multimodal learning, personalization, recommendation, and large language models.

Contact

For questions or updates, please contact jiahong.liu21@gmail.com, menglinyang@hkust-gz.edu.cn, or king@cse.cuhk.edu.hk.