Joint Binary Neural Network for Multi-label Learning with Applications to Emotion Classification

摘要: Recently the deep learning techniques have achieved success in multi-label classification due to its automatic representation ability and end-to-end framework. Existing neural networks can be divided into two kinds: binary relevance network (BRNN) threshold dependent (TDNN). However, former needs train a set of isolate which ignore dependencies between labels heavy computational load, while latter an additional function mechanism transform multi-class probabilities outputs. In this paper, we propose joint (JBNN), address these shortcomings. JBNN, text is fed logistic functions instead softmax function, multiple classifications are carried out synchronously one Moreover, relations captured via training on cross entropy (JBCE) loss. To better meet emotion classification, further proposed incorporate prior label JBCE The experimental results benchmark dataset show that our model performs significantly than state-of-the-art methods, both performance efficiency.