Thesis and Dissertation Workshop (WTDBD)
Complex Data, Metric Space, Temporal Evolution, Embedding, Similarity Search
In complex data, it is common to make similarity queries, using the features extracted of these data. These data are in general represented in metric spaces, where only the elements and their features are known. Considering the necessity of associate time to metric data, the objective is analyze the temporal evolution of the metric-temporal data, proposed on a previous work, through the embedding of these data to multidimensional spaces. In the multidimensional space, we can analyze the trajectories, estimate the data’s status in different moments of time and perform similarity search to this estimate in the multidimensional space. Initially, we intend to study algorithms of embedding that can preserve the distances between the elements as in the original space. Then, we will propose and evaluate different kinds of similarity search to perform estimates in the embedding space evaluating the outcomes in the moment of the search, starting with Range Query and Reverse k -NN. To finish, we intend to evaluate the use of multiple reference elements, when they are available, to calculate the estimates. With the results of the embedding and the queries on the estimates, we intend to validate the method proposed providing a support to real applications.
Isis Caroline O. V. de Sousa, Renato Bueno