A new learning algorithm for the hierarchical structure learning automata operating in the nonstationary S-model...

Area of Science:

• Reinforcement Learning
• Artificial Intelligence
• Control Theory

Background:

• Traditional reinforcement learning algorithms struggle in nonstationary environments where reward functions change over time.
• The relative reward strength algorithm offers a framework for adapting to changing conditions.
• Further enhancements are needed to guarantee convergence and improve performance.

Purpose of the Study:

• To propose an extended algorithm for relative reward strength.
• To demonstrate the algorithm's convergence properties in nonstationary environments.
• To validate the algorithm's effectiveness through computer simulations.

Main Methods:

• Development of an extended version of the relative reward strength algorithm.
• Theoretical analysis to prove convergence with probability 1 to the optimal path.
• Computer simulations to evaluate performance under nonstationary conditions.

Main Results:

• The proposed extended algorithm guarantees convergence to the optimal path with probability 1.
• Simulations confirm the algorithm's effectiveness in nonstationary environments.
• The enhanced algorithm demonstrates robust performance compared to existing methods.

Conclusions:

• The extended relative reward strength algorithm is a viable solution for optimal path finding in nonstationary environments.
• The theoretical guarantees and simulation results support the algorithm's practical applicability.
• This work contributes to the advancement of adaptive decision-making in dynamic artificial intelligence systems.