No abstract

In this paper, we tackle the problem of RGB-D semantic segmentation of indoor images. We take advantage of deconvolutional networks which can predict pixel-wise class labels, and develop a new structure for deconvolution of multiple modalities. We propose a novel feature transformation network to bridge the convolutional networks and deconvolutional networks. In the feature transformation network, we correlate the two modalities by discovering common features between them, as well as characterize each modality by discovering modality specific features. With the common features, we not only closely correlate the two modalities, but also allow them to borrow features from each other to enhance the representation of shared information. With specific features, we capture the visual patterns that are only visible in one modality. The proposed network achieves competitive segmentation accuracy on NYU depth dataset V1 and V2.

Distributed stochastic gradient descent (SGD) algorithms are widely deployed in training large-scale deep learning models, while the communication overhead among workers becomes the new system bottleneck. Recently proposed gradient sparsification techniques, especially Top-k sparsification with error compensation (TopK-SGD), can significantly reduce the communication traffic without obvious impact on the model accuracy. Some theoretical studies have been carried out to analyze the convergence property of TopK-SGD. However, existing studies do not dive into the details of Top-k operator in gradient sparsification and use relaxed bounds (e.g., exact bound of Random-k) for analysis; hence the derived results cannot well describe the real convergence performance of TopK-SGD. To this end, we first study the gradient distributions of TopK-SGD during training process through extensive experiments. We then theoretically derive a tighter bound for the Top-k operator. Finally, we exploit the property of gradient distribution to propose an approximate top-k selection algorithm, which is computing-efficient for GPUs, to improve the scaling efficiency of TopK-SGD by significantly reducing the computing overhead. Codes are available at: https://github.com/hclhkbu/GaussianK-SGD.

Abstract. The development of microprocessors design has been shifting to multi-core architectures. Therefore, it is expected that parallelism will play a significant role in future generations of applications. Throughout the years, there has been a myriad number of parallel programming models proposed. In choosing a parallel programming model, not only the performance aspect is important, but also qualitative the aspect of how well parallelism is abstracted to developers. A model with a well abstraction of parallelism leads to a higher application-development productivity. In this paper, we propose seven criteria to qualitatively evaluate parallel programming models. Our focus is on how parallelism is abstracted and presented to application developers. As a case study, we use these criteria to investigate six well-known parallel programming models in the HPC community.

Purpose Recently, various Twitter Sentiment Analysis (TSA) techniques have been developed, but little has paid attention to the microblogging feature – emojis, and few works have been conducted on the multi-class sentiment analysis of tweets. The purpose of this paper is to consider the popularity of emojis on Twitter and investigate the feasibility of an emoji training heuristic for multi-class sentiment classification of tweets. Tweets from the “2016 Orlando nightclub shooting” were used as a source of study. Besides, this study also aims to demonstrate how mapping can contribute to interpreting sentiments. Design/methodology/approach The authors presented a methodological framework to collect, pre-process, analyse and map public Twitter postings related to the shooting. The authors designed and implemented an emoji training heuristic, which automatically prepares the training data set, a feature needed in Big Data research. The authors improved upon the previous framework by advancing the pre-processing techniques, enhancing feature engineering and optimising the classification models. The authors constructed the sentiment model with a logistic regression classifier and selected features. Finally, the authors presented how to visualise citizen sentiments on maps dynamically using Mapbox. Findings The sentiment model constructed with the automatically annotated training sets using an emoji approach and selected features performs well in classifying tweets into five different sentiment classes, with a macro-averaged F-measure of 0.635, a macro-averaged accuracy of 0.689 and the MAEM of 0.530. Compared to those experimental results in related works, the results are satisfactory, indicating the model is effective and the proposed emoji training heuristic is useful and feasible in multi-class TSA. The maps authors created, provide a much easier-to-understand visual representation of the data, and make it more efficient to monitor citizen sentiments and distributions. Originality/value This work appears to be the first to conduct multi-class sentiment classification on Twitter with automatic annotation of training sets using emojis. Little attention has been paid to applying TSA to monitor the public’s attitudes towards terror attacks and country’s gun policies, the authors consider this work to be a pioneering work. Besides, the authors have introduced a new data set of 2016 Orlando Shooting tweets, which will be made available for other researchers to mine the public’s political opinions about gun policies.