Stochastic learning of multi-instance dictionary for earth mover’s distance-based histogram comparison

Abstract

Dictionary plays an important role in multi-instance data representation. It maps bags of instances to histograms. Earth mover's distance (EMD) is the most effective histogram distance metric for the application of multiinstance retrieval. However, up to now, there is no existing multi-instance dictionary learning methods designed for EMD based histogram comparison. To fill this gap, we develop the first EMD-optimal dictionary learning method using stochastic optimization method. In the stochastic learning framework, we have one triplet of bags, including one basic bag, one positive bag, and one negative bag. These bags are mapped to histograms using a multi-instance dictionary. We argue that the EMD between the basic histogram and the positive histogram should be smaller than that between the basic histogram and the negative histogram. Base on this condition, we design a hinge loss. By minimizing this hinge loss and some regularization terms of the dictionary, we update the dictionary instances. The experiments over multi-instance retrieval applications shows its effectiveness when compared to other dictionary learning methods over the problems of medical image retrieval and natural language relation classification.

A Novel Image Tag Completion Method Based on Convolutional Neural Transformation

“…The experiments of the problems of image annotation and image retrieval based on image tag completion over three benchmark data sets show the advantage of the proposed method. In the future, we will extend our work of CNN model to other machine learning problems beside image tag completion, such as computer vision [16,5,30,40,22,38,39], material engineering [32,33], portfolio choices [26,25,27,24,28], and biomedical engineering [4,3,2,1,21,13,10,23,12,31,29,9].…”

Bearing fault diagnosis based on variational autoencoder and non-local block wide kernel convolutional neural network

“…Data distribution similarity measure. The correlation between the generated data distribution and the original data distribution is assessed using the following metrics: Euclidean distance (ED), 35 KL divergence (KLD), 36 Cosine similarity (CS), 37 Earth Mover's Distance (EMD) 38 and Pearson correlation coefficient (PCC). 39 These five metrics are often used to measure the similarity between two data distributions.…”

Cross-model convolutional neural network for multiple modality data representation

“…For example, in the natural language problems, each sentence is treated as a sequence of words, and each word is represented by a word embedding vector. The sequence of word embedding vectors can be represented further by a CNN model for the problem of sematic classification [11,31,9]. Moreover, in computer vision applications, a video is also composed of a sequence of image frames, and we can also extract visual feature vector from each frame.…”