Related Experiment Videos
Automata with hierarchical control and evolutionary learning
Bio Systems
|January 1, 1988
Summary
We introduce a framework for hierarchical control using automata theory, enabling a universal automaton to emulate any other automaton on a graph. This framework defines automaton depth and aids in learning natural encodings for graph-based tasks.
Area of Science:
- Theoretical Computer Science
- Automata Theory
- Control Theory
Background:
- Hierarchical control systems are crucial for complex tasks.
- Automata theory provides formal models for computation and control.
- Graph traversal algorithms are fundamental in computer science.
Purpose of the Study:
- To develop an automata-theoretical framework for structured hierarchical control.
- To introduce a universal automaton capable of emulating other automata on graphs.
- To define and explore the concept of automaton depth based on hierarchical structure.
Main Methods:
- Proposing a framework based on rules and meta-rules for hierarchical control.
- Defining a universal hierarchically structured automaton (μ).
- Developing input mapping for automaton emulation and encoding.
- Analyzing efficient encodings of graph-search algorithms.
Main Results:
- A universal automaton (μ) can emulate any automaton on a given graph via input encoding.
- Efficient encodings correlate with the natural hierarchical structure of algorithms.
- A precise notion of automaton "depth" is established.
- Application to stochastic neural networks for learning structural encodings.
Conclusions:
- The proposed framework offers a formal approach to hierarchical control and automaton emulation.
- Learned encodings reflect the inherent hierarchical structure of graph-based algorithms.
- This work has implications for structural pattern recognition and learning systems.