Showing papers by "Joydeep Ghosh published in 2012"

Discovering important people and objects for egocentric video summarization

Abstract: We present a video summarization approach for egocentric or “wearable” camera data. Given hours of video, the proposed method produces a compact storyboard summary of the camera wearer's day. In contrast to traditional keyframe selection techniques, the resulting summary focuses on the most important objects and people with which the camera wearer interacts. To accomplish this, we develop region cues indicative of high-level saliency in egocentric video — such as the nearness to hands, gaze, and frequency of occurrence — and learn a regressor to predict the relative importance of any new region based on these cues. Using these predictions and a simple form of temporal event detection, our method selects frames for the storyboard that reflect the key object-driven happenings. Critically, the approach is neither camera-wearer-specific nor object-specific; that means the learned importance metric need not be trained for a given user or context, and it can predict the importance of objects and people that have never been seen previously. Our results with 17 hours of egocentric data show the method's promise relative to existing techniques for saliency and summarization.

Blind Image Quality Assessment Without Human Training Using Latent Quality Factors

Abstract: We propose a highly unsupervised, training free, no reference image quality assessment (IQA) model that is based on the hypothesis that distorted images have certain latent characteristics that differ from those of “natural” or “pristine” images. These latent characteristics are uncovered by applying a “topic model” to visual words extracted from an assortment of pristine and distorted images. For the latent characteristics to be discriminatory between pristine and distorted images, the choice of the visual words is important. We extract quality-aware visual words that are based on natural scene statistic features [1]. We show that the similarity between the probability of occurrence of the different topics in an unseen image and the distribution of latent topics averaged over a large number of pristine natural images yields a quality measure. This measure correlates well with human difference mean opinion scores on the LIVE IQA database [2].

EPIC: Efficient prediction of IC manufacturing hotspots with a unified meta-classification formulation

Abstract: In this paper we present EPIC, an efficient and effective predictor for IC manufacturing hotspots in deep sub-wavelength lithography. EPIC proposes a unified framework to combine different hotspot detection methods together, such as machine learning and pattern matching, using mathematical programming/optimization. EPIC algorithm has been tested on a number of industry benchmarks under advanced manufacturing conditions. It demonstrates so far the best capability in selectively combining the desirable features of various hotspot detection methods (3.5–8.2% accuracy improvement) as well as significant suppression of the detection noise (e.g., 80% false-alarm reduction). These characteristics make EPIC very suitable for conducting high performance physical verification and guiding efficient manufacturability friendly physical design.

Review quality aware collaborative filtering

Abstract: Probabilistic matrix factorization (PMF) and other popular approaches to collaborative filtering assume that the ratings given by users for products are genuine, and hence they give equal importance to all available ratings. However, this is not always true due to several reasons including the presence of opinion spam in product reviews. In this paper, the possibility of performing collaborative filtering while attaching weights or quality scores to the ratings is explored. The quality scores, which are determined from the corresponding review data are used to "up-weight" or "down-weight" the importance given to the individual rating while performing collaborative filtering, thereby improving the accuracy of the predictions. First, the measure used to capture the quality of the ratings is described. Different approaches for estimating the quality score based on the available review information are examined. Subsequently, a mathematical formulation to incorporate quality scores as weights for the ratings in the basic PMF framework is derived. Experimental evaluation on two product categories of a benchmark data set from Amazon.com demonstrates the efficacy of our approach.

An imputation-enhanced algorithm for ICU mortality prediction

Abstract: ICU patients are vulnerable to in-ICU morbidities and mortality, making accurate systems for identifying at-risk patients a necessity for improving clinical care. Here, we present an improved model for predicting in-hospital mortality using data collected from the first 48 hours of a patient's ICU stay. We generated predictive features for each patient using demographic data, the number of observations for each of 37 time-varying variables in hours 0–48 and 47–48 of the stay, and the last observed value for each variable. Missing data are a common problem in clinical data, and we therefore imputed missing values using the mean value for a patient's age and gender group. After imputing the missing data, we trained a logistic regression using this feature set. We evaluated model performance using the two metrics from the 2012 PhysioNet/CinC Challenge; the first measured model accuracy using the minimum of sensitivity and positive predictive value (Event 1), and the second measured model calibration using the Hosmer-Lemeshow H statistic (Event 2). Our model obtained Event 1 and 2 scores of 0.516 and 14.4 for test set B and 0.482 and 51.7 for test set C, respectively, providing better estimates of in-hospital mortality risk than existing methods such as SAPS-I.

Learning to Rank With Bregman Divergences and Monotone Retargeting

Abstract: This paper introduces a novel approach for learning to rank (LETOR) based on the notion of monotone retargeting. It involves minimizing a divergence between all monotonic increasing transformations of the training scores and a parameterized prediction function. The minimization is both over the transformations as well as over the parameters. It is applied to Bregman divergences, a large class of "distance like" functions that were recently shown to be the unique class that is statistically consistent with the normalized discounted gain (NDCG) criterion [19]. The algorithm uses alternating projection style updates, in which one set of simultaneous projections can be computed independent of the Bregman divergence and the other reduces to parameter estimation of a generalized linear model. This results in easily implemented, efficiently parallelizable algorithm for the LETOR task that enjoys global optimum guarantees under mild conditions. We present empirical results on benchmark datasets showing that this approach can outperform the state of the art NDCG consistent techniques.

Dating Texts without Explicit Temporal Cues

Abstract: This paper tackles temporal resolution of documents, such as determining when a document is about or when it was written, based only on its text. We apply techniques from information retrieval that predict dates via language models over a discretized timeline. Unlike most previous works, we rely {\it solely} on temporal cues implicit in the text. We consider both document-likelihood and divergence based techniques and several smoothing methods for both of them. Our best model predicts the mid-point of individuals' lives with a median of 22 and mean error of 36 years for Wikipedia biographies from 3800 B.C. to the present day. We also show that this approach works well when training on such biographies and predicting dates both for non-biographical Wikipedia pages about specific years (500 B.C. to 2010 A.D.) and for publication dates of short stories (1798 to 2008). Together, our work shows that, even in absence of temporal extraction resources, it is possible to achieve remarkable temporal locality across a diverse set of texts.

Learning to rank with Bregman divergences and monotone retargeting

Abstract: This paper introduces a novel approach for learning to rank (LETOR) based on the notion of monotone retargeting. It involves minimizing a divergence between all monotonic increasing transformations of the training scores and a parameterized prediction function. The minimization is both over the transformations as well as over the parameters. It is applied to Bregman divergences, a large class of "distance like" functions that were recently shown to be the unique class that is statistically consistent with the normalized discounted gain (NDCG) criterion [19]. The algorithm uses alternating projection style updates, in which one set of simultaneous projections can be computed independent of the Bregman divergence and the other reduces to parameter estimation of a generalized linear model. This results in easily implemented, efficiently parallelizable algorithm for the LETOR task that enjoys global optimum guarantees under mild conditions. We present empirical results on benchmark datasets showing that this approach can outperform the state of the art NDCG consistent techniques.

A probabilistic imputation framework for predictive analysis using variably aggregated, multi-source healthcare data

Abstract: In healthcare-related studies, individual patient or hospital data are not often publicly available due to privacy restrictions, legal issues or reporting norms. However, such measures may be provided at a higher or more aggregated level, such as state-level, county-level summaries or averages over health zones (HRR1 or HSA2). Such levels constitute partitions of the underlying individual level data, which may not match the data segments that would have been obtained if one clustered individual-level data. Treating these aggregated values as representatives for the individuals can result in the ecological fallacy. How can one run data mining procedures on such data where different variables are available at different levels of aggregation or granularity? We examine this problem in a clustering setting given a mix of individual-level and (arbitrarily) aggregated-level data. For this setting, a generative process of such data is constructed using a Bayesian directed graphical model. This model is further developed to capture the properties of the aggregated-level data using the Central Limit theorem. The model provides reasonable cluster centroids under certain conditions, and is extended to estimate the masked individual values for the aggregated data. The model parameters are learned using an approximated Gibbs sampling method, which employs the Metropolis-Hastings algorithm efficiently. A deterministic approximation algorithm is derived from the model, which scales up to massive datasets. Furthermore, the imputed features can help to improve the performance in subsequent predictive modeling tasks. Experimental results using data from the Dartmouth Health Atlas, CDC, and the U.S. Census Bureau are provided to illustrate the generality and capabilities of the proposed framework.

Transfer Learning with Cluster Ensembles.

Abstract: Traditional supervised learning algorithms typically assume that the training data and test data come from a common underlying distribution. Therefore, they are challenged by the mismatch between training and test distributions encountered in transfer learning situations. The problem is further exacerbated when the test data actually comes from a different domain and contains no labeled example. This paper describes an optimization framework that takes as input one or more classifiers learned on the source domain as well as the results of a cluster ensemble operating solely on the target domain, and yields a consensus labeling of the data in the target domain. This framework is fairly general in that it admits a wide range of loss functions and classification/clustering methods. Empirical results on both text and hyperspectral data indicate that the proposed method can yield superior classification results compared to applying certain other transductive and transfer learning techniques or naively applying the classifier (ensemble) learnt on the source domain to the target domain.

Probabilistic Combination of Classifier and Cluster Ensembles for Non-transductive Learning

Abstract: Unsupervised models can provide supplementary soft constraints to help classify new target data under the assumption that similar objects in the target set are more likely to share the same class label. Such models can also help detect possible differences between training and target distributions, which is useful in applications where concept drift may take place. This paper describes a Bayesian framework that takes as input class labels from existing classifiers (designed based on labeled data from the source domain), as well as cluster labels from a cluster ensemble operating solely on the target data to be classified, and yields a consensus labeling of the target data. This framework is particularly useful when the statistics of the target data drift or change from those of the training data. We also show that the proposed framework is privacy-aware and allows performing distributed learning when data/models have sharing restrictions. Experiments show that our framework can yield superior results to those provided by applying classifier ensembles only.

An Optimization Framework for Semi-Supervised and Transfer Learning using Multiple Classifiers and Clusterers

Abstract: Unsupervised models can provide supplementary soft constraints to help classify new, "target" data since similar instances in the target set are more likely to share the same class label. Such models can also help detect possible differences between training and target distributions, which is useful in applications where concept drift may take place, as in transfer learning settings. This paper describes a general optimization framework that takes as input class membership estimates from existing classifiers learnt on previously encountered "source" data, as well as a similarity matrix from a cluster ensemble operating solely on the target data to be classified, and yields a consensus labeling of the target data. This framework admits a wide range of loss functions and classification/clustering methods. It exploits properties of Bregman divergences in conjunction with Legendre duality to yield a principled and scalable approach. A variety of experiments show that the proposed framework can yield results substantially superior to those provided by popular transductive learning techniques or by naively applying classifiers learnt on the original task to the target data.

Probabilistic Combination of Classifier and Cluster Ensembles for Non-transductive Learning

Abstract: Unsupervised models can provide supplementary soft constraints to help classify new target data under the assumption that similar objects in the target set are more likely to share the same class label. Such models can also help detect possible differences between training and target distributions, which is useful in applications where concept drift may take place. This paper describes a Bayesian framework that takes as input class labels from existing classifiers (designed based on labeled data from the source domain), as well as cluster labels from a cluster ensemble operating solely on the target data to be classified, and yields a consensus labeling of the target data. This framework is particularly useful when the statistics of the target data drift or change from those of the training data. We also show that the proposed framework is privacy-aware and allows performing distributed learning when data/models have sharing restrictions. Experiments show that our framework can yield superior results to those provided by applying classifier ensembles only.

A Privacy-Aware Bayesian Approach for Combining Classifier and Cluster Ensembles

Abstract: This paper introduces a privacy-aware Bayesian approach that combines ensembles of classifiers and clusterers to perform semi-supervised and transductive learning. We consider scenarios where instances and their classification/clustering results are distributed across different data sites and have sharing restrictions. As a special case, the privacy aware computation of the model when instances of the target data are distributed across different data sites, is also discussed. Experimental results show that the proposed approach can provide good classification accuracies while adhering to the data/model sharing constraints.

Unnaturalness Modeling of Image Distortions

Abstract: Natural scene statistic (NSS) models are effective tools for formulating models of early visual processing. One area where NSS models have been successful is predicting human responses to image distortions, or image quality assessment (IQA) by quantifying unnaturalness introduced by distortions. Recent Blind IQA models use NSS features to form predictions of human judgments of distorted image quality without having available corresponding undistorted reference images. Successful learning blind models have previously been developed that learn to accurately predict human opinions of image quality by training them on databases of distorted images and associated human opinion scores. We introduce new NSS feature based blind IQA models that require even less information to attain good results. If human opinion scores of distorted images are not available, but a database of distorted images is, then opinion-less blind IQA models can be created that perform well. We have also found it possible to design blind IQA models without any source of prior information other than a database of distortionless " exemplar " images. An algorithm derived from such a completely blind model has only the distorted image to be quality-assessed available. Our new blind IQA models (Fig. 1) follow four processing steps (Fig. 2). Images are decomposed by an energy compacting filter bank then divisive normalized, yielding responses well-modeled as NSS. Either NSS features alone, or both NSS and distorted image statistic (DSS) features are used to create distributions of visual words. Quality prediction is expressed in terms of the Kullback-Leibler divergence between the distributions of visual words from distorted images and from the space of exemplar images. Both opinion blind and completely blind models compete well with standard non-blind metrics such as mean squared error (MSE) when tested on a large public IQA database (Tables 1 and 2).