Raven's Progressive Matrices (RPM) Agent
AI agent using knowledge-based techniques to solve visual analogy problems from Raven's Progressive Matrices, achieving 80% accuracy on advanced 3x3 matrices.
Designed an AI agent to solve Raven’s Progressive Matrices — a standard human intelligence test — using knowledge-based AI techniques. The agent processes visual analogy problems in both 2×2 and 3×3 formats and achieved 80 % accuracy on advanced problem sets.
Approach
- Implemented semantic networks to represent visual transformations between matrix frames
- Applied generate-and-test strategies to identify the best-fit answer from candidate options
- Modeled human-like reasoning processes, reflecting on how cognition can be captured computationally
Tools & Techniques
- Language: Python
- Course: Knowledge-Based AI (CS 7637), Georgia Tech
- Period: May – August 2024