Modeling outer products of features for image classification
Peng Qi, Shuochen Su, Xiaolin Hu · 2013
Recent studies have shown that sparse coding is an efficient method for feature quantization in image classification tasks. However, sparse coding can only capture linear statistical regularities among the features. In the paper, we show that features can be quantized in a nonlinear way by modeling their outer products. Experiments on some public datasets show that the proposed method can achieve comparable or better results than sparse coding.