4D effects are physical effects simulated in sync with videos, movies, and games to augment the events occurring in a story or a virtual world. Types of 4D effects commonly used for the immersive media may include seat motion, vibration, flash, wind, water, scent, thunderstorm, snow, and fog. Currently, the recognition of physical effects from a video is mainly conducted by human experts. Although 4D effects are promising in giving immersive experience and entertainment, this manual production has been the main obstacle to faster and wider application of 4D effects. In this paper, we utilize pretrained models of Convolutional Neural Networks (CNNs) to extract local visual features and propose a new representation method that combines extracted features into video level features. Classification tasks are conducted by employing Support Vector Machine (SVM). Comprehensive experiments are performed to investigate different architecture of CNNs and different type of features for 4D effect classification task and compare baseline average pooling method with our proposed video level representation. Our framework outperforms the baseline up to 2-3% in terms of mean average precision (mAP).

MOTIVATION: Quantitative mass spectrometry-based proteomics requires protein-level estimates and associated confidence measures. Challenges include the presence of low quality or incorrectly identified peptides and informative missingness. Furthermore, models are required for rolling peptide-level information up to the protein level. RESULTS: We present a statistical model that carefully accounts for informative missingness in peak intensities and allows unbiased, model-based, protein-level estimation and inference. The model is applicable to both label-based and label-free quantitation experiments. We also provide automated, model-based, algorithms for filtering of proteins and peptides as well as imputation of missing values. Two LC/MS datasets are used to illustrate the methods. In simulation studies, our methods are shown to achieve substantially more discoveries than standard alternatives. AVAILABILITY: The software has been made available in the open-source proteomics platform DAnTE (http://omics.pnl.gov/software/).

Trajectory segmentation is the process of partitioning a given trajectory into a small number of homogeneous segments w.r.t. some criteria. Conventional segmentation techniques only focus on the spatial features of the movement and could lead to spatially homogeneous segments but with presumably dissimilar temporal structures. Furthermore, trajectories could be over-segmented in the presence of outliers. In this paper, we propose a family of three trajectory segmentation methods that takes into account both geospatial and temporal structures of movement for the segmentation and is also robust with respect to time-referenced spatial outliers. The effectiveness of our methods is empirically demonstrated over three real-world datasets.

Feature subset selection (FSS) is a known technique to preprocess the data before performing any data mining tasks, e.g., classification and clustering. FSS provides both cost-effective predictors and a better understanding of the underlying process that generated the data. We propose a family of novel unsupervised methods for feature subset selection from multivariate time series (MTS) based on common principal component analysis, termed CLeVer. Traditional FSS techniques, such as recursive feature elimination (RFE) and Fisher criterion (FC), have been applied to MTS data sets, e.g., brain computer interface (BCI) data sets. However, these techniques may lose the correlation information among features, while our proposed techniques utilize the properties of the principal component analysis to retain that information. In order to evaluate the effectiveness of our selected subset of features, we employ classification as the target data mining task. Our exhaustive experiments show that CLeVer outperforms RFE, FC, and random selection by up to a factor of two in terms of the classification accuracy, while taking up to 2 orders of magnitude less processing time than RFE and FC.