Determining optimal action in consideration of risk

A computer readable storage medium tangibly embodying a computer readable program code having computer readable instructions which when implemented, cause a computer to carry out the steps comprising:

• (a) selecting a state from possible states in a current term, wherein said current term belongs to a time horizon;
• (b) selecting an action from at least one action candidate that can be executed in a selected state;
• (c) calculating a probability distribution of an evaluation value for a selected action, wherein said evaluation value corresponds to a reward obtained in said current term and an optimal value for next term;
• (d) calculating a risk measure using said probability distribution of said evaluation value;
• (e) determining a weighting function conforming to at least one preference by taking said risk measure into consideration;
• (f) calculating a value measure of said selected action;
• (g) repeating steps (b) through (f) for all other said action candidates that can be executed in said selected state;
• and
• (h) comparing said value measures of said selected actions in order to determine an optimal action for said selected state.

A system that determines an optimal action in consideration of risk, the system comprising:

• a first selection module for selecting a state from possible states in a current term, wherein said current term belongs to a time horizon;
• a second selection module for selecting an action from action candidates that can be executed in a selected state;
• a probability distribution module for calculating a probability distribution of an evaluation value for a selected action, wherein said evaluation value corresponds to a reward obtained in said current term and an optimal value for next term;
• a risk measure module for calculating a risk measure using said probability distribution of said evaluation value;
• a decision module for determining a weighting function conforming to a preference by taking said risk measure into consideration;
• a value measure module for calculating a value measure of said selected action;
• a first control module for repeating the steps from said first selection module, said second selection module, said probability distribution module, said risk measure module, said decision module, and said value measure module for all other said action candidates that can be executed in said selected state;
• and
• a comparing module for comparing said value measures of said selected actions in order to determine an optimal action for said selected state.