Latent Semantic Representation Learning for Scene Classification
Xin Li, Yuhong Guo · 2014
The performance of machine learning methods is heavily dependent on the choice of data represen-tation. In real world applications such as scene recognition problems, the widely used low-level input features can fail to explain the high-level semantic label concepts. In this work, we ad-dress this problem by proposing a novel patch-based latent variable model to integrate latent contextual representation learning and classifi-cation model training in one joint optimization framework. Within this framework, the latent layer of variables bridge the gap between inputs and outputs by providing discriminative expla-nations for the semantic output labels, while be-ing predictable from the low-level input features. Experiments conducted on standard scene recog-nition tasks demonstrate the efficacy of the pro-posed approach, comparing to the state-of-the-art scene recognition methods. 1.