Adversarial examples, referred to as augmented data points generated by imperceptible perturbations of input samples, have recently drawn much attention. Well-crafted adversarial examples may even mislead state-of-the-art deep neural network (DNN) models to make wrong predictions easily. To alleviate this problem, many studies have focused on investigating how adversarial examples can be generated and/or effectively handled. All existing works tackle this problem in the Euclidean space. In this paper, we extend the learning of adversarial examples to the more general Riemannian space over DNNs. The proposed work is important in that (1) it is a generalized learning methodology since Riemmanian space will be degraded to the Euclidean space in a special case; (2) it is the first work to tackle the adversarial example problem tractably through the perspective of Riemannian geometry; (3) from the perspective of geometry, our method leads to the steepest direction of the loss function, by considering the second order information of the loss function. We also provide a theoretical study showing that our proposed method can truly find the descent direction for the loss function, with a comparable computational time against traditional adversarial methods. Finally, the proposed framework demonstrates superior performance over traditional counterpart methods, using benchmark data including MNIST, CIFAR-10 and SVHN.

Technology has enabled anyone with an Internet connection to easily create and share their ideas, opinions and content with millions of other people around the world. Much of the content being posted and consumed online is multimodal. With billions of phones, tablets and PCs shipping today with built-in cameras and a host of new video-equipped wearables like Google Glass on the horizon, the amount of video on the Internet will only continue to increase. It has become increasingly difficult for researchers to keep up with this deluge of multimodal content, let alone organize or make sense of it. Mining useful knowledge from video is a critical need that will grow exponentially, in pace with the global growth of content. This is particularly important in sentiment analysis, as both service and product reviews are gradually shifting from unimodal to multimodal. We present a novel method to extract features from visual and textual modalities using deep convolutional neural networks. By feeding such features to a multiple kernel learning classifier, we significantly outperform the state of the art of multimodal emotion recognition and sentiment analysis on different datasets.

Text localization in born-digital images is usually performed using methods designed for scene text images. Based on the observation that text strokes in born-digital images mostly have complete contours and the pixels on the contours have high contrast compared with the adjacent non-text pixels, we propose a method to extract candidate text components using local contrast. First, the image is segmented into smooth and non-smooth regions. After removing non-text smooth regions, the remaining smooth regions are merged with non-smooth regions to form a candidate text image, which is binarized into high-value and low-value connected components (CCs). The CCs undergo CC filtering, line grouping and line classification to give the text localization result. Experimental results on the born-digital dataset of ICDAR2013 robust reading competition demonstrate the efficiency and superiority of the proposed method.