Goal Based Agents
Goal based agents have some kind of a goal that hey have to work for, and it is not really measured.
Before goal based agents, model goals were fixed and implicitly defined from the conditions.
Utility based agents
Here the agent has some kind of a measure, maybe we can call it a happiness level, and it is a number, and it represents how much will the agent be satisfied if it took a certain action. The agent here will try to maximize the utility function. Here I define the rules.
Utility cab be considered as multiple goals
Learning based agent modules
These type of agents have a machine learning kind of style, We have the critic block which can be considered as the loss function. We also have learning element block which we can think of having the weights.
The problem generator is a bit of a weird block, what it does is - according to the doctor - it does simulations taking into consideration what the agent has learned and reflects it in the performance element
