Abstract:
실시예는 인공 지능 추론 장치 및 방법을 개시한다. 실시예는 인공 지능 추론 방법으로, 학습된 신경망 및 파라미터를 학습 프레임워크에 독립적인 상위 레벨 언어의 실행 코드로 변환하는 단계와, 실행 코드를 가속 연산 필요 여부에 따라 범용 언어(General Purpose Language, GPL) 코드와 도메인 특정 언어(Domain Specific Language, DSL) 코드로 분리하는 단계 및 분리된 GPL 코드 및 DSL 코드를 하드웨어에 최적화된 타겟 코드로 생성하는 단계를 포함할 수 있다.
Abstract:
The present invention relates to a position calculation apparatus and a position calculation method thereof which separate reception intensity of irregular communication into three layers by using two sensing nodes in a defense sensor network environment with restricted resources, and calculate the position of an object for each layer more accurately. The position calculation method includes the steps of: determining the grade of a first reception intensity corresponding to an object communicating with a first node; calculating a distance from the first node of the object on the basis of the grade of the first reception intensity; determining the grade of a second reception intensity corresponding to an object communicating with a second node; calculating a distance from the second node of the object on the basis of the grade of the second reception intensity; calculating two estimated positions on the basis of the distance from the first node and the distance from the second node; and determining one of the two estimated position as a final position of the object by predicting the movement of the object. [Reference numerals] (100) Communication part; (200) First grade determining part; (300) First estimated dot line calculation part; (400) Second grade determining part; (500) Second estimated dot line calculation part; (600) Estimated position calculation part; (700) Final position calculation part