Multiple Instance Learning

Multiple Instance Learning (MIL) is a weakly supervised learning algorithm where the training data is organized in bags, each containing a set of instances $X=\{x_1,x_2,\ldots,x_M\}$, and each bag has a single label $Y \in \{0, 1\}$. The algorithm assumes that each instance within a bag has its own label $y_1, y_2, \ldots, y_M$, but these labels are unknown during the training process. The standard Multiple Instance Learning assumption posits that if all instances in a bag are negative, then the bag is negative; if at least one instance in the bag is positive, then the bag is positive. This algorithm has significant advantages in handling complex data structures, particularly applicable to fields such as medical image classification.