Audio Set classification with attention model: A probabilistic perspective

Abstract: This paper investigates the classification of the Audio Set dataset. Audio Set is a large scale weakly labelled dataset of sound clips. Previous work used multiple instance learning (MIL) to classify weakly labelled data. In MIL, a bag consists of several instances, and a bag is labelled positive if at least one instances in the audio clip is positive. A bag is labelled negative if all the instances in the bag are negative. We propose an attention model to tackle the MIL problem and explain this attention model from a novel probabilistic perspective. We define a probability space on each bag, where each instance in the bag has a trainable probability measure for each class. Then the classification of a bag is the expectation of the classification output of the instances in the bag with respect to the learned probability measure. Experimental results show that our proposed attention model modeled by fully connected deep neural network obtains mAP of 0.327 on Audio Set dataset, outperforming the Google's baseline of 0.314 and recurrent neural network of 0.325.