According to @1982marrVision, any system carrying out an information processing task must be understood at three levels: computational theory, representation and algorithm, and hardware implementation.
| Level | What theories at this level are about | Characteristic questions, after Marr |
|---|---|---|
| Computational | A system’s goals and/or problems it is trying to solve. Principles that the system can use to produce outputs from available inputs in natural environments. | ”What is the goal of the computation, why is it appropriate, and what is the logic of the strategy by which it can be carried out?” |
| Algorithmic | The procedures executed by a system to produce a solution. The representations or data structures over which transformations operate. | ”How can this computational theory be implemented? In particular, what is the representation for the input and output, and what is the algorithm for the transformation?” |
| Implementational | How algorithms and data structures are instantiated in physical systems. | ”How can the representation and algorithm be realized physically?” |
Related notes:
- Levels of cognitive theories
- Levels of granularity in computational systems
- Top-down vs. bottom-up epistemological strategies in cognitive science
References
- @1982marrVision, Vision
- @2024griffithsBayesian, Bayesian Models of Cognition