Research on image sentiment analysis technology based on sparse representation

• DOI: 10.1049/cit2.12074
• Publication Date: 2022-01-05T03:34:32Z
• Authors: Xiaofang Jin; Yinan Wu; Ying Xu; Chang Sun
• Keywords: Representation; Singular value; Value (mathematics); Feature (linguistics)

Many methods based on deep learning have achieved amazing results in image sentiment analysis. However, these existing usually pursue high accuracy, ignoring the effect model training efficiency. Considering that when faced with large-scale analysis tasks, accuracy rate often requires long experimental time. In view of weakness, a method can greatly improve efficiency only small fluctuations is proposed, and singular value decomposition (SVD) used to find sparse feature image, which are vectors strong discriminativeness effectively reduce redundant information; The authors propose Fast Dictionary Learning algorithm (FDL), combine neural network representation. This K-Singular Value Decomposition, through iteration, it calculation time case fluctuation accuracy. Moreover, effectiveness proposed evaluated FER2013 dataset. By adding decomposition, test suite increased by 0.53%, total experiment was shortened 8.2%; 36.3%.