The trade-off in RL between trying new actions (exploration) and using known profitable actions (exploitation). In trading terms, an agent may explore unfamiliar strategies or stick to historically profitable ones. Effective RL balances exploration (to discover new profitable patterns) with exploitation (to use current knowledge). This dilemma is crucial: too much exploration wastes time, too little may miss better strategies. Techniques like epsilon-greedy or decaying exploration rates manage this balance.

Back to Robots Glossary Index