The new mathematical model can teach AI agents to make the right decisions
Björn Lindenberg has published a new dissertation in mathematics demonstrating the application of reinforcement learning in AI to develop efficient techniques for autonomous decision-making in varied situations. Reward systems can be created to encourage desirable behaviour, such as managing robots and network traffic or determining the best pricing schemes for financial products.
A digital decision-maker, known as an agent, learns to make decisions by interaction with its environment and getting rewards or penalties based on how successfully it executes its activities. Reinforcement learning is a component of AI.
rewards and penalties are used to reinforce learning by behaving in a context and receiving feedback based on that behaviour. AI gradually learns to carry out desired activities and enhance its performance in the job at hand by maximising rewards and minimising penalties.