Physical Sciences and Mathematics Commons^™

Facial Expression Recognition By De-Expression Residue Learning, Huiyuan Yang, Umur Ciftci, Lijun Yin

Computer Science Faculty Research & Creative Works

A facial expression is a combination of an expressive component and a neutral component of a person. In this paper, we propose to recognize facial expressions by extracting information of the expressive component through a de-expression learning procedure, called De-expression Residue Learning (DeRL). First, a generative model is trained by cGAN. This model generates the corresponding neutral face image for any input face image. We call this procedure de-expression because the expressive information is filtered out by the generative model; however, the expressive information is still recorded in the intermediate layers. Given the neutral face image, unlike previous works using …

Improving Error-Bounded Compression For Cosmological Simulation, Sihuan Li, Sheng Di, Xin Liang, Zizhong Chen, Franck Cappello

Computer Science Faculty Research & Creative Works

Cosmological simulations may produce extremely large amount of data, such that its successful run depends on large storage capacity and huge I/O bandwidth, especially in the exascale computing scale. Effective error-bounded lossy compressors with both high compression ratios and low data distortion can significantly reduce the total data size while guaranteeing the data valid for post-analysis. In this poster, we propose a novel, efficient compression model for cosmological N-body simulation framework, by combining the advantages of both space-based compression and time-based compression. The evaluation with a well-known cosmological simulation code shows that our proposed solution can get much higher compression …

Exploring Best Lossy Compression Strategy By Combining Sz With Spatiotemporal Decimation, Xin Liang, Sheng Di, Sihuan Li, Dingwen Tao, Zizhong Chen, Franck Cappello

Computer Science Faculty Research & Creative Works

In today’s extreme-scale scientific simulations, vast volumes of data are being produced such that the data cannot be accommodated by the parallel file system or the data writing/ reading performance will be fairly low because of limited I/O bandwidth. In the past decade, many snapshot-based (or space-based) lossy compressors have been developed, most of which rely on the smoothness of the data in space. However, the simulation data may get more and more complicated in space over time steps, such that the compression ratios decrease significantly. In this paper, we propose a novel, hybrid lossy compression method by leveraging spatiotemporal …

Multicellular Models Bridging Intracellular Signaling And Gene Transcription To Population Dynamics, Mohammad Aminul Islam, Satyaki Roy, Sajal K. Das, Dipak Barua

Computer Science Faculty Research & Creative Works

Cell signaling and gene transcription occur at faster time scales compared to cellular death, division, and evolution. Bridging these multiscale events in a model is computationally challenging. We introduce a framework for the systematic development of multiscale cell population models. Using message passing interface (MPI) parallelism, the framework creates a population model from a single-cell biochemical network model. It launches parallel simulations on a single-cell model and treats each stand-alone parallel process as a cell object. MPI mediates cell-to-cell and cell-to-environment communications in a server-client fashion. In the framework, model-specific higher level rules link the intracellular molecular events to cellular …

Phase Contrast Time-Lapse Microscopy Datasets With Automated And Manual Cell Tracking Annotations, Dai Fei Elmer Ker, Zhaozheng Yin, For Full List Of Authors, See Publisher's Website.

Computer Science Faculty Research & Creative Works

Phase contrast time-lapse microscopy is a non-destructive technique that generates large volumes of image-based information to quantify the behaviour of individual cells or cell populations. To guide the development of algorithms for computer-aided cell tracking and analysis, 48 time-lapse image sequences, each spanning approximately 3.5 days, were generated with accompanying ground truths for C2C12 myoblast cells cultured under 4 different media conditions, including with fibroblast growth factor 2 (FGF2), bone morphogenetic protein 2 (BMP2), FGF2 + BMP2, and control (no growth factor). The ground truths generated contain information for tracking at least 3 parent cells and their descendants within these …

A Distributed Semi-Supervised Platform For Dnase-Seq Data Analytics Using Deep Generative Convolutional Networks, Shayan Shams, Richard Platania, Joohyun Kim, Jian Zhang, Kisung Lee, Seungwon Yang, Seung Jong Park

Computer Science Faculty Research & Creative Works

A deep learning approach for analyzing DNase-seq datasets is presented, which has promising potentials for unraveling biological underpinnings on transcription regulation mechanisms. Further understanding of these mechanisms can lead to important advances in life sciences in general and drug, biomarker discovery, and cancer research in particular. Motivated by recent remarkable advances in the field of deep learning, we developed a platform, Deep Semi-Supervised DNase-seq Analytics (DSSDA). Primarily empowered by deep generative Convolutional Networks (ConvNets), the most notable aspect is the capability of semi-supervised learning, which is highly beneficial for common biological settings often plagued with a less sufficient number of …

Gpu-Accelerated Large-Scale Genome Assembly, Sayan Goswami, Kisung Lee, Shayan Shams, Seung Jong Park

Computer Science Faculty Research & Creative Works

Spurred by a widening gap between hardware accelerators and traditional processors, numerous bioinformatics applications have harnessed the computing power of GPUS and reported substantial performance improvements compared to their CPU-based counterparts. However, most of these GPU-based applications only focus on the read alignment problem, while the field of de novo assembly still relies mostly on CPU-based solutions. This is primarily due to the nature of the assembly workload which is not only compute-intensive but also extremely data-intensive. Such workloads require large memories, making it difficult to adapt them to use GPUS with their limited memory capacities. To the best of …

Towards Distributed Cyberinfrastructure For Smart Cities Using Big Data And Deep Learning Technologies, Shayan Shams, Sayan Goswami, Kisung Lee, Seungwon Yang, Seung Jong Park

Computer Science Faculty Research & Creative Works

Recent advances in big data and deep learning technologies have enabled researchers across many disciplines to gain new insight into large and complex data. For example, deep neural networks are being widely used to analyze various types of data including images, videos, texts, and time-series data. In another example, various disciplines such as sociology, social work, and criminology are analyzing crowd-sourced and online social network data using big data technologies to gain new insight from a plethora of data. Even though many different types of data are being generated and analyzed in various domains, the development of distributed city-level cyberinfrastructure …

Automated Design Of Network Security Metrics, Aaron Scott Pope, Daniel R. Tauritz, Robert Morning, Alexander D. Kent

Computer Science Faculty Research & Creative Works

Many abstract security measurements are based on characteristics of a graph that represents the network. These are typically simple and quick to compute but are often of little practical use in making real-world predictions. Practical network security is often measured using simulation or real-world exercises. These approaches better represent realistic outcomes but can be costly and time-consuming. This work aims to combine the strengths of these two approaches, developing efficient heuristics that accurately predict attack success. Hyper-heuristic machine learning techniques, trained on network attack simulation training data, are used to produce novel graph-based security metrics. These low-cost metrics serve as …

The Automated Design Of Probabilistic Selection Methods For Evolutionary Algorithms, Samuel N. Richter, Daniel R. Tauritz

Computer Science Faculty Research & Creative Works

Selection functions enable Evolutionary Algorithms (EAs) to apply selection pressure to a population of individuals, by regulating the probability that an individual's genes survive, typically based on fitness. Various conventional fitness based selection methods exist, each providing a unique relationship between the fitnesses of individuals in a population and their chances of selection. However, the full space of selection algorithms is only limited by max algorithm size, and each possible selection algorithm is optimal for some EA configuration applied to a particular problem class. Therefore, improved performance may be expected by tuning an EA's selection algorithm to the problem at …

Evolution Of Network Enumeration Strategies In Emulated Computer Networks, Sean Harris, Eric Michalak, Kevin Schoonover, Adam Gausmann, Hannah Reinbolt, Joshua Herman, Daniel R. Tauritz, Chris Rawlings, Aaron Scott Pope

Computer Science Faculty Research & Creative Works

Successful attacks on computer networks today do not often owe their victory to directly overcoming strong security measures set up by the defender. Rather, most attacks succeed because the number of possible vulnerabilities are too large for humans to fully protect without making a mistake. Regardless of the security elsewhere, a skilled attacker can exploit a single vulnerability in a defensive system and negate the benefits of those security measures. This paper presents an evolutionary framework for evolving attacker agents in a real, emulated network environment using genetic programming, as a foundation for coevolutionary systems which can automatically discover and …

Improving Performance Of Iterative Methods By Lossy Checkponting, Dingwen Tao, Sheng Di, Xin Liang, Zizhong Chen, Franck Cappello

Computer Science Faculty Research & Creative Works

Iterative methods are commonly used approaches to solve large, sparse linear systems, which are fundamental operations for many modern scientific simulations. When the large-scale iterative methods are running with a large number of ranks in parallel, they have to checkpoint the dynamic variables periodically in case of unavoidable fail-stop errors, requiring fast I/O systems and large storage space. To this end, significantly reducing the checkpointing overhead is critical to improving the overall performance of iterative methods. Our contribution is fourfold. (1) We propose a novel lossy checkpointing scheme that can significantly improve the checkpointing performance of iterative methods by leveraging …

Identity-Adaptive Facial Expression Recognition Through Expression Regeneration Using Conditional Generative Adversarial Networks, Huiyuan Yang, Zheng Zhang, Lijun Yin

Computer Science Faculty Research & Creative Works

Subject variation is a challenging issue for facial expression recognition, especially when handling unseen subjects with small-scale labeled facial expression databases. Although transfer learning has been widely used to tackle the problem, the performance degrades on new data. In this paper, we present a novel approach (so-called IA-gen) to alleviate the issue of subject variations by regenerating expressions from any input facial images. First of all, we train conditional generative models to generate six prototypic facial expressions from any given query face image while keeping the identity related information unchanged. Generative Adversarial Networks are employed to train the conditional generative …

Heterogeneous Activity Causes A Nonlinear Increase In The Group Energy Use Of Ant Workers Isolated From Queen And Brood, Nolan Ferral, Kyara Holloway, Mingzhong Li, Zhaozheng Yin, Chen Hou

Computer Science Faculty Research & Creative Works

Increasing evidence has shown that the energy use of ant colonies increases sublinearly with colony size so that large colonies consume less per capita energy than small colonies. It has been postulated that social environment (e.g., in the presence of queen and brood) is critical for the sublinear group energetics, and a few studies of ant workers isolated from queens and brood observed linear relationships between group energetics and size. In this paper, we hypothesize that the sublinear energetics arise from the heterogeneity of activity in ant groups, that is, large groups have relatively more inactive members than small groups. …

Software Engineering: Guest Editor's Introduction, Bruce M. Mcmillin

Computer Science Faculty Research & Creative Works

Engineering complex systems requires an extensive technical and semantic knowledge. Because of their complexity, components of these systems can interact in unpredictable ways -- sequentially and concurrently….

The three papers in this theme issue have the potential to address the design challenges of complex systems spanning multiple domains and interacting in potentially unexpected ways. The knowledge required goes beyond basic computing concepts and design, to include an examination of interactions and multiple aspects of the application domain. We hope you find these articles interesting and insightful.