One of the most important machine learning methods is supervised learning in which labled data is used to traine the system.But some times determining the label of each individual data sample is difficult and even impossible, and only the lable of a set of samples is specified. In these situations a kind of leaning called multi – instance learning is used, which is introduced for working with ambiguous data and the possibility of kashida; TEXT-ALIGN: justify; LINE-HEIGHT: 25.2pt; TEXT-KASHIDA: 0%; MARGIN: 0cm 0cm 0pt; unicode-bidi: embed; DIRECTION: ltr"