Hyuk-Ro Park

MPIRace-check: detection of message races in MPI programs

Message races, which can cause nondeterministic executions of a parallel program, should be detected for debugging because nondeterminism makes debugging parallel programs a difficult task. Even though there are some tools to detect message races in MPI programs, they do not provide practical information to locate and debug message races in MPI programs. In this paper, we present an on-the-fly detection tool, which is MPIRace-Check, for debugging MPI programs written in C language. MPIRace-Check detects and reports all race conditions in all processes by checking the concurrency of the communication between processes. Also it reports the message races with some practical information such as the line number of a source code, the processes number, and the channel information which are involved in the races. By providing those information, it lets programmers distinguish of unintended races among the reported races, and lets the programmers know directly where the races occur in a huge source code. In the experiment we will show that MPIRace-Check detects the races using some testing programs as well as the tool is efficient.

Automatically improving image quality using tensor voting

A novel corrupted region detection technique based on tensor voting is proposed to automatically improve the image quality. This method is suitable for restoring degraded images and enhancing binary images. First, the input images are converted into layered images in which each layer contains objects having similar characteristics. By encoding the pixels in the layered images with second-order tensors and performing voting among them, the corrupted regions are automatically detected using the resulting tensors. These corrupted regions are then restored to improve the image quality. The experimental results obtained from automatic image restoration and binary image enhancement applications show that our method can successfully detect and correct the corrupted regions.

SVD-LDA: A Combined Model for Text Classification

Abstract: Text data has always accounted for a major portion of the world’s information. As the volume of information increases exponentially, the portion of text data also increases significantly. Text classification is therefore still an important area of research. LDA is an updated, probabilistic model which has been used in many applications in many other fields. As regards text data, LDA also has many applications, which has been applied various enhancements. However, it seems that no applications take care of the input for LDA. In this paper, we suggest a way to map the input space to a reduced space, which may avoid the unreliability, ambiguity and redundancy of individual terms as descriptors. The purpose of this paper is to show that LDA can be perfectly performed in a “clean and clear” space. Experiments are conducted on 20 News Groups data sets. The results show that the proposed method can boost the classification results when the appropriate choice of rank of the reduced space is determined.

Texture Comparison with an Orientation Matching Scheme

Abstract —Texture is an important visual feature for image analysis. Many approaches have been proposed to model and analyze texture features. Although these approaches significantly contribute to various image-based applications, most of these methods are sensitive to the changes in the scale and orientation of the texture pattern. Because textures vary in scale and orientations frequently, this easily leads to pattern mismatching if the features are compared to each other without considering the scale and/or orientation of textures. This paper suggests an Orientation Matching Scheme (OMS) to ease the problem of mismatching rotated patterns. In OMS, a pair of texture features will be compared to each other at various orientations to identify the best matched direction for comparison. A database including rotated texture images was generated for experiments. A synthetic retrieving experiment was conducted on the generated database to examine the performance of the proposed scheme. We also applied OMS to the similarity computation in a K-means clustering algorithm. The purpose of using K-means is to examine the scheme exhaustively in unpromising conditions, where initialized seeds are randomly selected and algorithms work heuristically. Results from both types of experiments show that the proposed OMS can help improve the performance when dealing with rotated patterns.

Constructing an Evaluation Set for Korean Sentiment Analysis Systems Incorporating the Category and the Strength of Sentiment

Sentiment analysis is concerned with extracting and analyzing different kinds of user sentiment expressed in a variety of social media such as blog and twitter. Although sentiment analysis techniques are actively studied for these days, evaluation sets are not developed yet for Korean sentiment analysis. In this paper, we constructed an evaluation set for Korean sentiment analysis. To evaluate sentiment analysis systems more throughly, each sentence in our evaluation set is tagged with the polarity of the sentiment as well as the category and the strength of the sentiment. We divide kinds of sentiment into 7 positive categories and 15 negative categories. Each category is given the strength of the sentiment from 1 to 3. Our evaluation set consists of 3,270 sentences extracted from various social media. For each sentence, 5 human taggers assigned the category and the strength of the sentiment expressed in the sentence. The ratio of inter-taggers agreement was 93% in the polarity, 70% in the category, 58% in the strength of sentiment. The ratio of inter-taggers agreement our evaluation set is a bit higher than other evaluation sets developed for German and Spanish. This result shows our evaluation set can be used as a reliable resource for the evaluation of sentiment analysis systems.

Verb Sense Disambiguation using Subordinating Case Information

Homographs can have multiple senses. In order to understand the meaning of a sentence, it is necessary to identify which sense isused for each word in the sentence. Previous researches on this problem heavily relied on the word co-occurrence information. However, we noticed that in case of verbs, information about subordinating cases of verbs can be utilized to further improve the performance of word sense disambiguation. Different senses require different sets of subordinating cases. In this paper, we propose the verb sense disambiguation using subordinating case information. The case information acquire postposition features in Standard Korean Dictionary. Our experiment on 12 high-frequency verb homographs shows that adding case information can improve the performance of word sense disambiguation by 1.34%, from 97.3% to 98.7%. The amount of improvement may seem marginal, we think it is meaningful because the error ratio reduced to less than a half, from 2.7% to 1.3%.

Obtaining Object by Using Optimal Threshold for Saliency Map Thresholding

Salient object attracts more and more attention from researchers due to its important role in many fields of multimedia processing like tracking, segmentation, adaptive compression, and content-base image retrieval. Usually, a saliency map is binarized into black and white map, which is considered as the binary mask of the salient object in the image. Still, the threshold is heuristically chosen or parametrically controlled. This paper suggests using the global optimal threshold to perform saliency map thresholding. This work also considers the usage of multi-level optimal thresholds and the local adaptive thresholds in the experiments. These experimental results show that using global optimal threshold method is better than parametric controlled or local adaptive threshold method.

Automatic Image Restoration Based on Tensor Voting

An automatic image restoration method is proposed for text images despite severe occlusion and noise. 3D tensor voting framework is used to analyze surface areas to detect corrupted regions. These corrupted regions are then restored by an adaptive median filter or image completing. The experimental results attained from several text images show that good images can be achieved from degraded ones by using the proposed method.

Improving the Performance of Web Search Using Users' Bookmarks

In this paper we propose using the bookmark and combination with the other feature for web search engine. As the majority of current web search engines use keyword-based method for similarity computing, they cannot discriminate important web pages among huge amount of search results. The proposed system reflects collaborative evaluation by users by accumulating the number of bookmark on a web page. In this paper we implement a user interface that helps to store bookmark count and click count separately according to pages by multiple users, and estimate a ranking function though several features on returned result page. The experimental results show that the accuracy of proposed system is improved by as much as 30.5% compared to conventional web search engines.

Visualization of Affect-Relations of Message Races for Debugging MPI Programs

Detecting unaffected races is important for debugging MPI parallel programs, because unaffected races can cause the occurrence of affected races which do not need to be debugged. However, the previous techniques can not discern unaffected races from affected races so that programmers will be easily overwhelmed by the vast information of race detection. In this paper, we present a new visualization which lets programmers know which race is affected or not. For this, our technique checks whether any message racing toward a race is affected or not based on happen- before relation, and also checks which process influences a race during an execution. After the execution, it visualizes the affect-relations of the detected races. Therefore, our visualization helps for programmers to effectively distinguish unaffected races from affected races, and to debug MPI parallel programs.