Introduction to knowledge graphs
Knowledge representation is one of the open problems of AI and has very ancient roots (Leibniz believed that the whole knowledge could be represented and used to conduct calculations). The interest in knowledge representation is based on the fact that it represents the first step in conducting computer reasoning. Once this knowledge is organized in an orderly manner, it can be used to design inference algorithms and solve reasoning problems. Early studies focused on using deduction to solve problems about organized entities (e.g., through the use of ontologies). This has worked well for many toy problems, but it is laborious, often requires a whole set of hardcoded rules, and risks succumbing to combinatorial explosion. Because search in these spaces could be extremely computationally expensive, an attempt was made to define two concepts:
- Limited rationality: Finding a solution but also considering the cost of it
- Heuristic search: Limiting...
