Labelling Unknown Class Samples

In a practical world, any living being survives through life long learning process. In computation, generic machine learning models are close-set recognition oriented. To create life long machine learning model, algorithms need to deal with the open-set recognition process. In this work, we have proposed a framework to recognize unknown samples of an unknown class and label them with an appropriate class label. We design decision criteria based on the knowledge of the existing two known classes and applied them to the pool of samples containing unknown samples. Experimental results show that our approach provides similar accuracy as the normal closed-set classification process, and thus, it establishes the correctness of our approach.