Relighting and Viewpoint Control Input (b) Head Pose Control (c) Shadow Diffusion Figure 1. Holo-... more Relighting and Viewpoint Control Input (b) Head Pose Control (c) Shadow Diffusion Figure 1. Holo-Relighting performs volumetric relighting on a single input portrait image, allowing users to individually control (1) lighting effect via an environment map, (2) camera viewpoint and (3) head pose. It is highly expressive and can render complex illumination effects on in-the-wild human faces with accurate view consistency (a ). Controls are well disentangled to produce a realistic rendering of moving shadows cast by a point light while rotating the head (b). Practical photographic applications such as shadow diffusion (softening) are also made feasible with our system (c).
reference image source image Photoswap reference image source image Photoswap Figure 1: Photoswap... more reference image source image Photoswap reference image source image Photoswap Figure 1: Photoswap can effortlessly replace the subject in a source image, which could be either synthetic (first two rows) or real (bottom row), with a personalized subject specified in reference images, while preserving the original subject pose and the composition of the source image.
Figure 1. LightPainter is an interactive lighting editing system that takes in an input image wit... more Figure 1. LightPainter is an interactive lighting editing system that takes in an input image with freehand scribbles drawn on top and renders the correspondingly relit portrait. It enables creative portrait lighting editing (left) and allows users to reproduce a target lighting effect with ease (right).
A Binary Stock Event Model for stock trends forecasting: Forecasting stock trends via a simple and accurate approach with machine learning
2011 11th International Conference on Intelligent Systems Design and Applications, 2011
ABSTRACT The volatile and stochastic characteristics of securities make it challenging to predict... more ABSTRACT The volatile and stochastic characteristics of securities make it challenging to predict even tomorrow's stock prices. Better estimation of stock trends can be accomplished using both the significant and well-constructed set of features. Moreover, the prediction capability will gain momentum as we build the right model to capture unobservable attributes of the varying tendencies. In this paper, we propose a Binary Stock Event Model (BSEM) and generate features sets based on it in order to better predict the future trends of the stock market. We apply two learning models such as a Bayesian Naive Classifier and a Support Vector Machine to prove the efficiency of our approach in the aspects of prediction accuracy and computational cost. Our experiments demonstrate that the prediction accuracies are around 70–80% in one day predictions. In addition, our back-testing proves that our trading model outperforms well-known technical indicator based trading strategies with regards to cumulative returns by 30%–100%. As a result, this paper suggests that our BSEM based stock forecasting shows its excellence with regards to prediction accuracy and cumulative returns in a real world dataset.
While recent work has shown that a worker’s performance can be more accurately modeled by tempora... more While recent work has shown that a worker’s performance can be more accurately modeled by temporal correlation in task performance, a fundamental challenge remains in the need for expert gold labels to evaluate a worker’s performance. To solve this problem, we explore two methods of utilizing limited gold labels, initial training and periodic updating. Furthermore, we present a novel way of learning a prediction model in the absence of gold labels with uncertaintyaware learning and soft-label updating. Our experiment with a real crowdsourcing dataset demonstrates that periodic updating tends to show better performance than initial training when the number of gold labels are very limited (< 25).
Proceedings of the 23rd International Conference on World Wide Web, 2014
We investigate a method of crowdsourced task routing based on matrix factorization. From a prelim... more We investigate a method of crowdsourced task routing based on matrix factorization. From a preliminary analysis of a real crowdsourced data, we begin an exploration of how to route crowdsourcing task via Matrix factorization (MF) which efficiently estimate missing values in a worker-task matrix. Our preliminary results show the benefits of task routing over random assignment, the strength of probabilistic MF over baseline methods.
View-centric Operational Transformation for Collaborative Editing
2006 International Conference on Collaborative Computing: Networking, Applications and Worksharing, 2006
ABSTRACT Collaborative editing enables multiple users who reside remotely to share and edit some ... more ABSTRACT Collaborative editing enables multiple users who reside remotely to share and edit some documents at the same time. It is fundamentally based on operational transformation which adjusts the position of operation according to the transformed execution order. For a last decade, many researches have been performed in this area and the correctness and possibility of operational transformation have been proved. Even though operational transformation gives us the possibility of collaborative editing, it has a limitation with a viewpoint of usability and efficiency. In other words, the existing operational transformation is devised without considering the properties of collaborative editing such as the frequency of operational transformation and human-centric viewpoint. In this paper, we would like to introduce view-centric operational transformation which considers the priority of transforming operation according to user&#39;s viewpoint. Using this way, we have tried to improve the existing operational transformation and provide more useful and efficient collaborative editing
While temporal behavioral patterns can be discerned to un-derlie real crowd work, prior studies h... more While temporal behavioral patterns can be discerned to un-derlie real crowd work, prior studies have typically modeled worker performance under a simplified i.i.d. assumption. To better model such temporal worker behavior, we propose a time-series label prediction model for crowd work. This latent variable model captures and summarizes past worker behav-ior, enabling us to better predict the quality of each worker's next label. Given inherent uncertainty in prediction, we also investigate a decision reject option to balance the tradeoff between prediction accuracy vs. coverage. Results show our model improves accuracy of both label prediction on real crowd worker data, as well as data quality overall.
Flexible authentication and authorization architecture for grid computing
2005 International Conference on Parallel Processing Workshops (ICPPW'05), 2005
... Proxy Certificates for Dynamic Delegation”, In Annual PKI workshop, Von Welch, Siebenlis, Ian... more ... Proxy Certificates for Dynamic Delegation”, In Annual PKI workshop, Von Welch, Siebenlis, Ian Foster, John Bresnaban, Karl Czajkowski, Jarek Gawor, Carl Kesselman, Sam Meder, Laura Pearlman and Steven Tuecke , “Secu-...
ABSTRACT With the advent of grid computing, multiple users can make use of heterogeneous computin... more ABSTRACT With the advent of grid computing, multiple users can make use of heterogeneous computing nodes and data nodes for collaborative research. Hence, collaboration tool is inevitably included in grid computing application. In this paper, we propose the collaborative editor for multiple user’s real-time research. And we describes how collaborative editor can be easily installed and executed.
Proceedings of the 35th international ACM SIGIR conference on Research and development in information retrieval - SIGIR '12, 2012
In crowdsourced relevance judging, each crowd worker typically judges only a small number of exam... more In crowdsourced relevance judging, each crowd worker typically judges only a small number of examples, yielding a sparse and imbalanced set of judgments in which relatively few workers influence output consensus labels, particularly with simple consensus methods like majority voting. We show how probabilistic matrix factorization, a standard approach in collaborative filtering, can be used to infer missing worker judgments such that all workers influence output labels. Given complete worker judgments inferred by PMF, we evaluate impact in unsupervised and supervised scenarios. In the supervised case, we consider both weighted voting and worker selection strategies based on worker accuracy. Experiments on crowd judgments from the 2010 TREC Relevance Feedback Track show promise of the PMF approach merits further investigation and analysis.
We describe our submission to the Image Relevance Assessment Task (IRAT) at the 2012 Text REtriev... more We describe our submission to the Image Relevance Assessment Task (IRAT) at the 2012 Text REtrieval Conference (TREC) Crowdsourcing Track. Four aspects distinguish our approach: 1) an interface for cohesive, efficient topicbased relevance judging and reporting judgment confidence; 2) a variant of Welinder and Perona's method for online crowdsourcing (inferring quality of the judgments and judges during data collection in order to dynamically optimize data collection); 3) a completely unsupervised approach using no labeled data for either training or tuning; and 4) automatic generation of individualized error reports for each crowd worker, supporting transparent assessment and education of workers. Our system was built start-to-finish in two weeks, and we collected approximately 44,000 labels for about $40 US.
This paper considers the challenge of evaluating a set of classifiers, as done in shared task eva... more This paper considers the challenge of evaluating a set of classifiers, as done in shared task evaluations like the KDD Cup or NIST TREC, without expert labels. While expert labels provide the traditional cornerstone for evaluating statistical learners, limited or expensive access to experts represents a practical bottleneck. Instead, we seek methodology for estimating performance of the classifiers (relative and absolute) which is more scalable than expert labeling yet preserves high correlation with evaluation based on expert labels. We consider both: 1) using only labels automatically generated by the classifiers themselves (blind evaluation); and 2) using labels obtained via crowdsourcing. While crowdsourcing methods are lauded for scalability, using such data for evaluation raises serious concerns given the prevalence of label noise. In regard to blind evaluation, two broad strategies are investigated: combine & score and score & combine. Combine & Score methods infer a single "pseudo-gold" label set by aggregating classifier labels; classifiers are then evaluated based on this single pseudo-gold label set. On the other hand, score & combine methods: i) sample multiple label sets from classifier outputs, ii) evaluate classifiers on each label set, and iii) average classifier performance across label sets. When additional crowd labels are also collected, we investigate two alternative avenues for exploiting them: 1) direct evaluation of classifiers; or 2) supervision of combine-and-score methods. To assess generality of our techniques, classifier performance is measured using four common classification metrics, with statistical significance tests establishing relative performance of the classifiers for each metric. Finally, we measure both score and rank correlations between estimated classifier performance vs. actual performance according to expert judgments. Rigorous evaluation of classifiers from the TREC 2011 Crowdsourcing Track shows reliable evaluation can be achieved without reliance on expert labels.
Quality assurance in crowdsourced annotation often involves having a given example labeled multip... more Quality assurance in crowdsourced annotation often involves having a given example labeled multiple times by different workers, then aggregating these labels. Unfortunately, the worker-example label matrix is typically sparse and imbalanced for two reasons: 1) the average crowd worker judges few examples; and 2) few labels are typically collected per example to reduce cost. To address this missing data problem, we propose use of probabilistic matrix factorization (PMF), a standard approach in collaborative filtering. To evaluate our approach, we measure accuracy of consensus labels computed from the input sparse matrix vs. the PMF-inferred complete matrix. We consider both unsupervised and supervised settings. In the supervised case, we evaluate both weighted voting and worker selection. Experiments are performed on both a synthetic data set and a real data set: crowd relevance judgments taken from the 2010 NIST TREC Relevance Feedback Track.
Abstract. Generally, grid services are implemented and deployed with various kinds of components.... more Abstract. Generally, grid services are implemented and deployed with various kinds of components. Intrinsically, this type of grid service ar-chitecture has unessential components only to use the small parts of function which provides. Hence, it can be criticized by reason of its own property like heaviness, redundancy and low performance. In real usage of grid service, most of users would like to use a compact and flexible grid service because of convenience and easiness. In this paper, we have proposed flexible and compact grid service architecture for grid users. For the flexible service, we adopt a hierarchical structure and to achieve the property of compactness, we modularize our grid system architecture which can be deployed with existed Grid components. Conclusively, we show how to operate and implement our system. In addition, we present the result of experiments by using our system architecture with Medical Grid Service. 1
Diamond: A SPARQL Query Engine, for Linked Data Based on the Rete Match
This paper describes a system, Diamond, which uses the Rete Match algorithm to evaluate SPARQL qu... more This paper describes a system, Diamond, which uses the Rete Match algorithm to evaluate SPARQL queries on distributed RDF data in the Linked Data model. In the Linked Data model, as a query is being evaluated, additional linked data can be identified as additional data to be evaluated by the query; the process may re-peat indefinitely. Casting Linked Data query evaluation as a cyclic behavior enables making a constructive analogy with the behavior of a forward-chaining AI production system. The Rete match algorithm is the most commonly used implementation technique for AI pro-duction systems. Where AI production systems are known to make a relatively consistent number of changes to working memory per cycle, dereferencing URIs in the linked data model is a potentially volatile process. The paper provides an overview of Diamonds ar-chitectural elements that concern integrating the Rete match with the crawling of Linked Data and provides an overview of a set of Rete operators needed to...
Relighting and Viewpoint Control Input (b) Head Pose Control (c) Shadow Diffusion Figure 1. Holo-... more Relighting and Viewpoint Control Input (b) Head Pose Control (c) Shadow Diffusion Figure 1. Holo-Relighting performs volumetric relighting on a single input portrait image, allowing users to individually control (1) lighting effect via an environment map, (2) camera viewpoint and (3) head pose. It is highly expressive and can render complex illumination effects on in-the-wild human faces with accurate view consistency (a ). Controls are well disentangled to produce a realistic rendering of moving shadows cast by a point light while rotating the head (b). Practical photographic applications such as shadow diffusion (softening) are also made feasible with our system (c).
reference image source image Photoswap reference image source image Photoswap Figure 1: Photoswap... more reference image source image Photoswap reference image source image Photoswap Figure 1: Photoswap can effortlessly replace the subject in a source image, which could be either synthetic (first two rows) or real (bottom row), with a personalized subject specified in reference images, while preserving the original subject pose and the composition of the source image.
Figure 1. LightPainter is an interactive lighting editing system that takes in an input image wit... more Figure 1. LightPainter is an interactive lighting editing system that takes in an input image with freehand scribbles drawn on top and renders the correspondingly relit portrait. It enables creative portrait lighting editing (left) and allows users to reproduce a target lighting effect with ease (right).
A Binary Stock Event Model for stock trends forecasting: Forecasting stock trends via a simple and accurate approach with machine learning
2011 11th International Conference on Intelligent Systems Design and Applications, 2011
ABSTRACT The volatile and stochastic characteristics of securities make it challenging to predict... more ABSTRACT The volatile and stochastic characteristics of securities make it challenging to predict even tomorrow&#39;s stock prices. Better estimation of stock trends can be accomplished using both the significant and well-constructed set of features. Moreover, the prediction capability will gain momentum as we build the right model to capture unobservable attributes of the varying tendencies. In this paper, we propose a Binary Stock Event Model (BSEM) and generate features sets based on it in order to better predict the future trends of the stock market. We apply two learning models such as a Bayesian Naive Classifier and a Support Vector Machine to prove the efficiency of our approach in the aspects of prediction accuracy and computational cost. Our experiments demonstrate that the prediction accuracies are around 70–80% in one day predictions. In addition, our back-testing proves that our trading model outperforms well-known technical indicator based trading strategies with regards to cumulative returns by 30%–100%. As a result, this paper suggests that our BSEM based stock forecasting shows its excellence with regards to prediction accuracy and cumulative returns in a real world dataset.
While recent work has shown that a worker’s performance can be more accurately modeled by tempora... more While recent work has shown that a worker’s performance can be more accurately modeled by temporal correlation in task performance, a fundamental challenge remains in the need for expert gold labels to evaluate a worker’s performance. To solve this problem, we explore two methods of utilizing limited gold labels, initial training and periodic updating. Furthermore, we present a novel way of learning a prediction model in the absence of gold labels with uncertaintyaware learning and soft-label updating. Our experiment with a real crowdsourcing dataset demonstrates that periodic updating tends to show better performance than initial training when the number of gold labels are very limited (< 25).
Proceedings of the 23rd International Conference on World Wide Web, 2014
We investigate a method of crowdsourced task routing based on matrix factorization. From a prelim... more We investigate a method of crowdsourced task routing based on matrix factorization. From a preliminary analysis of a real crowdsourced data, we begin an exploration of how to route crowdsourcing task via Matrix factorization (MF) which efficiently estimate missing values in a worker-task matrix. Our preliminary results show the benefits of task routing over random assignment, the strength of probabilistic MF over baseline methods.
View-centric Operational Transformation for Collaborative Editing
2006 International Conference on Collaborative Computing: Networking, Applications and Worksharing, 2006
ABSTRACT Collaborative editing enables multiple users who reside remotely to share and edit some ... more ABSTRACT Collaborative editing enables multiple users who reside remotely to share and edit some documents at the same time. It is fundamentally based on operational transformation which adjusts the position of operation according to the transformed execution order. For a last decade, many researches have been performed in this area and the correctness and possibility of operational transformation have been proved. Even though operational transformation gives us the possibility of collaborative editing, it has a limitation with a viewpoint of usability and efficiency. In other words, the existing operational transformation is devised without considering the properties of collaborative editing such as the frequency of operational transformation and human-centric viewpoint. In this paper, we would like to introduce view-centric operational transformation which considers the priority of transforming operation according to user&#39;s viewpoint. Using this way, we have tried to improve the existing operational transformation and provide more useful and efficient collaborative editing
While temporal behavioral patterns can be discerned to un-derlie real crowd work, prior studies h... more While temporal behavioral patterns can be discerned to un-derlie real crowd work, prior studies have typically modeled worker performance under a simplified i.i.d. assumption. To better model such temporal worker behavior, we propose a time-series label prediction model for crowd work. This latent variable model captures and summarizes past worker behav-ior, enabling us to better predict the quality of each worker's next label. Given inherent uncertainty in prediction, we also investigate a decision reject option to balance the tradeoff between prediction accuracy vs. coverage. Results show our model improves accuracy of both label prediction on real crowd worker data, as well as data quality overall.
Flexible authentication and authorization architecture for grid computing
2005 International Conference on Parallel Processing Workshops (ICPPW'05), 2005
... Proxy Certificates for Dynamic Delegation”, In Annual PKI workshop, Von Welch, Siebenlis, Ian... more ... Proxy Certificates for Dynamic Delegation”, In Annual PKI workshop, Von Welch, Siebenlis, Ian Foster, John Bresnaban, Karl Czajkowski, Jarek Gawor, Carl Kesselman, Sam Meder, Laura Pearlman and Steven Tuecke , “Secu-...
ABSTRACT With the advent of grid computing, multiple users can make use of heterogeneous computin... more ABSTRACT With the advent of grid computing, multiple users can make use of heterogeneous computing nodes and data nodes for collaborative research. Hence, collaboration tool is inevitably included in grid computing application. In this paper, we propose the collaborative editor for multiple user’s real-time research. And we describes how collaborative editor can be easily installed and executed.
Proceedings of the 35th international ACM SIGIR conference on Research and development in information retrieval - SIGIR '12, 2012
In crowdsourced relevance judging, each crowd worker typically judges only a small number of exam... more In crowdsourced relevance judging, each crowd worker typically judges only a small number of examples, yielding a sparse and imbalanced set of judgments in which relatively few workers influence output consensus labels, particularly with simple consensus methods like majority voting. We show how probabilistic matrix factorization, a standard approach in collaborative filtering, can be used to infer missing worker judgments such that all workers influence output labels. Given complete worker judgments inferred by PMF, we evaluate impact in unsupervised and supervised scenarios. In the supervised case, we consider both weighted voting and worker selection strategies based on worker accuracy. Experiments on crowd judgments from the 2010 TREC Relevance Feedback Track show promise of the PMF approach merits further investigation and analysis.
We describe our submission to the Image Relevance Assessment Task (IRAT) at the 2012 Text REtriev... more We describe our submission to the Image Relevance Assessment Task (IRAT) at the 2012 Text REtrieval Conference (TREC) Crowdsourcing Track. Four aspects distinguish our approach: 1) an interface for cohesive, efficient topicbased relevance judging and reporting judgment confidence; 2) a variant of Welinder and Perona's method for online crowdsourcing (inferring quality of the judgments and judges during data collection in order to dynamically optimize data collection); 3) a completely unsupervised approach using no labeled data for either training or tuning; and 4) automatic generation of individualized error reports for each crowd worker, supporting transparent assessment and education of workers. Our system was built start-to-finish in two weeks, and we collected approximately 44,000 labels for about $40 US.
This paper considers the challenge of evaluating a set of classifiers, as done in shared task eva... more This paper considers the challenge of evaluating a set of classifiers, as done in shared task evaluations like the KDD Cup or NIST TREC, without expert labels. While expert labels provide the traditional cornerstone for evaluating statistical learners, limited or expensive access to experts represents a practical bottleneck. Instead, we seek methodology for estimating performance of the classifiers (relative and absolute) which is more scalable than expert labeling yet preserves high correlation with evaluation based on expert labels. We consider both: 1) using only labels automatically generated by the classifiers themselves (blind evaluation); and 2) using labels obtained via crowdsourcing. While crowdsourcing methods are lauded for scalability, using such data for evaluation raises serious concerns given the prevalence of label noise. In regard to blind evaluation, two broad strategies are investigated: combine & score and score & combine. Combine & Score methods infer a single "pseudo-gold" label set by aggregating classifier labels; classifiers are then evaluated based on this single pseudo-gold label set. On the other hand, score & combine methods: i) sample multiple label sets from classifier outputs, ii) evaluate classifiers on each label set, and iii) average classifier performance across label sets. When additional crowd labels are also collected, we investigate two alternative avenues for exploiting them: 1) direct evaluation of classifiers; or 2) supervision of combine-and-score methods. To assess generality of our techniques, classifier performance is measured using four common classification metrics, with statistical significance tests establishing relative performance of the classifiers for each metric. Finally, we measure both score and rank correlations between estimated classifier performance vs. actual performance according to expert judgments. Rigorous evaluation of classifiers from the TREC 2011 Crowdsourcing Track shows reliable evaluation can be achieved without reliance on expert labels.
Quality assurance in crowdsourced annotation often involves having a given example labeled multip... more Quality assurance in crowdsourced annotation often involves having a given example labeled multiple times by different workers, then aggregating these labels. Unfortunately, the worker-example label matrix is typically sparse and imbalanced for two reasons: 1) the average crowd worker judges few examples; and 2) few labels are typically collected per example to reduce cost. To address this missing data problem, we propose use of probabilistic matrix factorization (PMF), a standard approach in collaborative filtering. To evaluate our approach, we measure accuracy of consensus labels computed from the input sparse matrix vs. the PMF-inferred complete matrix. We consider both unsupervised and supervised settings. In the supervised case, we evaluate both weighted voting and worker selection. Experiments are performed on both a synthetic data set and a real data set: crowd relevance judgments taken from the 2010 NIST TREC Relevance Feedback Track.
Abstract. Generally, grid services are implemented and deployed with various kinds of components.... more Abstract. Generally, grid services are implemented and deployed with various kinds of components. Intrinsically, this type of grid service ar-chitecture has unessential components only to use the small parts of function which provides. Hence, it can be criticized by reason of its own property like heaviness, redundancy and low performance. In real usage of grid service, most of users would like to use a compact and flexible grid service because of convenience and easiness. In this paper, we have proposed flexible and compact grid service architecture for grid users. For the flexible service, we adopt a hierarchical structure and to achieve the property of compactness, we modularize our grid system architecture which can be deployed with existed Grid components. Conclusively, we show how to operate and implement our system. In addition, we present the result of experiments by using our system architecture with Medical Grid Service. 1
Diamond: A SPARQL Query Engine, for Linked Data Based on the Rete Match
This paper describes a system, Diamond, which uses the Rete Match algorithm to evaluate SPARQL qu... more This paper describes a system, Diamond, which uses the Rete Match algorithm to evaluate SPARQL queries on distributed RDF data in the Linked Data model. In the Linked Data model, as a query is being evaluated, additional linked data can be identified as additional data to be evaluated by the query; the process may re-peat indefinitely. Casting Linked Data query evaluation as a cyclic behavior enables making a constructive analogy with the behavior of a forward-chaining AI production system. The Rete match algorithm is the most commonly used implementation technique for AI pro-duction systems. Where AI production systems are known to make a relatively consistent number of changes to working memory per cycle, dereferencing URIs in the linked data model is a potentially volatile process. The paper provides an overview of Diamonds ar-chitectural elements that concern integrating the Rete match with the crawling of Linked Data and provides an overview of a set of Rete operators needed to...
Uploads
Papers by HyunJoon Jung