A partially supervised learning algorithm for linearly separable systems

An important aspect of human learning is the ability to select effective samples to learn and utilize the experience to infer the outcomes of new events. This type of learning is characterized as partially supervised learning. A learning algorithm of this type is suggested for linearly separable systems. The algorithm selects a subset S from a finite set X of linearly separable vectors to construct a linear classifier that can correctly classify all the vectors in X. The sample set S is chosen without any prior knowledge of how the vectors in X-S are classified. The computational complexity of the algorithm is analyzed, and the lower bound on the size of the sample set is established