PROJECT TITLE :

A k-Nearest Neighbor Multilabel Ranking Algorithm with Application to Content-based Image Retrieval - 2017

ABSTRACT:

Multilabel ranking is an important Machine Learning task with several applications, such as content-based mostly image retrieval (CBIR). But, when the amount of labels is giant, traditional algorithms are either infeasible or show poor performance. During this paper, we propose a easy nonetheless effective multilabel ranking algorithm that is primarily based on k-nearest neighbor paradigm. The proposed algorithm ranks labels consistent with the chances of the label association using the neighboring samples around a question sample. Totally different from traditional approaches, we take solely positive samples into consideration and determine the model parameters by directly optimizing ranking loss measures. We have a tendency to evaluated the proposed algorithm using four common multilabel datasets. The proposed algorithm achieves equivalent or higher performance than other instance-primarily based learning algorithms. When applied to a CBIR system with a dataset of 1 million samples and over one hundred ninety thousand labels, that is abundant larger than any other multilabel datasets used earlier, the proposed algorithm clearly outperforms the competing algorithms.