IMPROVED AUDIO CLASSIFICATION USING A NOVEL NON-LINEAR DIMENSIONALITY REDUCTION ENSEMBLE APPROACH

Two important categories of machine learning method-ologies have recently attracted much interest in classifica-tion research and its applications. On one side, unsuper-vised and semi-supervised learning allow to benefit from the availability of larger sets of training data, even if not fully annotated with class labels, and of larger sets of di-verse feature representations, through novel dimensional-ity reduction schemes. On the other side, ensemble meth-ods allow to benefit from more diversity in base learners though larger data and feature sets. In this paper, we pro-pose a novel ensemble learning approach making use of recent non-linear dimensionality reduction methods. More precisely, we apply t-SNE (t-distributed Stochastic Neigh-bor Embedding) to a large feature set to come up with em-beddings of various dimensionality. A k-NN classifier is then obtained for each embedding, leading to an ensemble whose estimates can then be combined, making use of var-ious ensemble combination rules from the literature. The rationale of this approach resides in its potential capacity to better handle manifolds of different dimensionality in different regions of the feature space. We evaluate the ap-proach on a transductive audio classification task, where only part of the whole data set is labeled. We confirm that dimensionality reduction by itself can improve perfor-mance (by 40 % relative), and that creating an ensemble through the proposed approach further reduces classifica-tion error rate by about 10 % relative. 1