We propose a simple approach which combines the strengths of probabilistic graphical models and deep learning architectures for solving the multi-label classification task, focusing specifically on image and video data. First, we show that the performance of previous approaches that combine Markov Random Fields with neural networks can be modestly improved by leveraging more powerful methods such as iterative join graph propagation, integer linear programming, and 1 regularizationbased structure learning. Then we propose a new modeling framework called deep dependency networks, which augments a dependency network, a model that is easy to train and learns more accurate dependencies but is limited to Gibbs sampling for inference, to the output layer of a neural network. We show that despite its simplicity, jointly learning this new architecture yields significant improvements in performance over the baseline neural network. In particular, our experimental evaluation on three video activity classification datasets: Charades, Textually Annotated Cooking Scenes (TACoS), and Wetlab, and three multi-label image classification datasets: MS-COCO, PASCAL VOC, and NUS-WIDE show that deep dependency networks are almost always superior to pure neural architectures that do not use dependency networks.
Measuring semantic similarity between words plays vital role in information retrieval and natural language processing. The existing system uses page counts and snippets retrieved by a search engine to measure semantic similarity between words. Various similarity scores are calculated from the page counts retrieved by the search engine for the queried conjunctive words. A lexical pattern extraction algorithm identifies the patterns from the snippets. Different patterns showing the same semantic relation are clustered using a lexical pattern clustering algorithm. The existing system makes use of Support Vector Machines to combine the similarity scores from page counts and clusters of patterns from snippets for measuring similarity. We propose a different machine learning approach called Latent Structural Support Vector Machine which can handle the missing data values which occurs frequently in statistical data analysis. The proposed system also makes a comparative study between similarity results from both SVM and LS-SVM.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.