LEARNING SYSTEM AND LEARNING METHOD

PROBLEM TO BE SOLVED: To achieve optimization of activities to be learnt connected in series to acquire the activity from a status.SOLUTION: A learning system that learns each activity to be learnt for a group of to-be-learnt objects formed by multiple to-be-learnt objects to acquire the activity from a status includes a first controller and a second controller. The second controller acquires a first learning result on each to-be-learnt object, transmits the first learning result to the first controller so as to give the first learning result of a given to-be-learnt object to another to-be-learnt object, acquires a second learning result of each to-be-learnt object acquired upon transmission of the first learning result to the first controller, and evaluates and outputs a group of to-be-learnt objects on the basis of each first learning result and of each second learning result. The first controller creates the status of another to-be-learnt object when data to which the activity of a given to-be-learnt object contribute is given to another to-be-learnt object with a delay, and selects the next activity of another learning object and transmits the learning result on the basis of the activity of the given to-be-learnt object and of the status of another to-be-learnt object.SELECTED DRAWING: Figure 1 【課題】状態から行動を得る直列接続された複数の学習対象の行動の最適化を図ること。【解決手段】状態から行動を得る複数の学習対象により構成された学習対象群について各学習対象の行動を学習する学習システムは、第１コントローラと第２コントローラを有し、第２コントローラは学習対象群の各々の第１学習結果を取得し、ある学習対象の第１学習結果を他の学習対象に与えるように第１コントローラに送信し、第１学習結果を第１コントローラに送信した結果得られる学習対象群の各々の第２学習結果を取得し、各第１学習結果と各第２学習結果に基づいて学習対象群を評価して出力し、第１コントローラは、ある学習対象の行動が寄与したデータが遅延を伴って他の学習対象に与えられることにより他の学習対象の状態を生成し、ある学習対象の行動と他の学習対象の状態とに基づいて、他の学習対象の次の行動を選択し学習結果として送信する。【選択図】図１