Graph Theory and Games: A Case Study on New York Times Connections
summary
Represents each of the 16 puzzle words as graph nodes and the four hidden groups as colors, stores embeddings and word associations in a Neo4j-backed workflow, uses Cypher and graph algorithms to compute coherence scores and guide choices, and trains a custom RGCNN with reinforcement learning on stratified train/test puzzles. Prototype ranks candidate groupings, tracks NMI grouping accuracy, reaches human-level on many boards but still trips on the hardest purple/ambiguous categories.
AI-written summary · about
chapters (9)
- 0:00Introduction and overview
- 2:08How Connections works
- 5:59Why Connections interests AI
- 9:30Graph coloring problem basics
- 11:15Quantifying puzzle difficulty
- 16:09Building a graph-based solver
- 18:35Reinforcement learning approach
- 23:10Why graph neural networks matter
- 24:13Q&A and closing
related talks
01h 41mGraph Intelligence: Enhance Reasoning and Retrieval Using Graph Analytics
46mGraph Algorithms: Predict Real-World Behavior
29mA Graph Entity Resolution Playbook
31mGraph Analytics - From the Whiteboard to Your Toolbox
32mUnderstanding Complexly Connected Data using Graph Databases
13mManaging Connections with Graphs: Uses and Challenges
06mUnderstanding Player Patterns by Combining Knowledge-Based Data Abstraction with Interactive ...
43mGraphs are Everywhere