Physical Sciences and Mathematics Commons^™

New Explainable Active Learning Approach For Recommender Systems, Sami Khenissi, Behnoush Abdollahi, Wenlong Sun, Pegah Sagheb, Olfa Nasraoui

Commonwealth Computational Summit

Introduction and Motivations

• Recommender Systems are intelligent programs that analyze patterns between items and users to predict the user’s taste.

Objective

• Design an efficient Active Learning Strategy to increase the explainability and the accuracy of an “Explainable Matrix Factorization” model.

A Network Tomography Approach For Traffic Monitoring In Smart Cities, Ruoxi Zhang, Sara Newman, Marco Ortolani, Simone Silvestri

Commonwealth Computational Summit

Various urban planning and managing activities required by a Smart City are feasible because of traffic monitoring. As such, this project proposes a network tomography-based approach that can be applied to road networks to achieve a cost-efficient, flexible, and scalable monitor deployment. Due to the algebraic approach of network tomography, the selection of monitoring intersections can be solved through the use of matrices, with its rows representing paths between two intersections, and its columns representing links in the road network. Because the goal of the algorithm is to provide an inexpensive monitor set, this problem can be translated into a …

Bio-Inspired Disaster Response Networks, Vijay K. Shah, Simone Silvestri, Sajal K. Das, Satyaki Roy

Commonwealth Computational Summit

Large-scale natural disasters (e.g., Earthquake, Hurricane) –

• Three times as many disasters between 1980 and 2016 compared to 1940-1980. (EM-DAT – The International Disaster Database)
• Since 1990, 217 million people affected each year. (The New England Journal of Medicine)

Aftermath a disaster,

• Loss of human lives and property
• Lack of food, clean drinking water, shelter etc.
• Disruption of infrastructure networks (e.g. cellular towers) and other public infrastructures (e.g. power sources) – Our focus !

Formation Of Supermassive Black Holes In The Early Universe: High-Resolution Numerical Simulations Of Radiation Transfer Inside Collapsing Gas, Yang Luo, Kazem Ardaneh, Isaac Shlosman, Kentaro Nagamine, John Wise, Mitchell C. Begelman

Commonwealth Computational Summit

Observations of high-redshift quasars reveal that super massive black holes (SMBHs) with masses exceeding 10⁹ M_⊙ formed as early as redshift z ~ 7 [1,3,6]. This means that SMBHs have already formed ~700 million years after the Big Bang. How did such SMBHs could grow so quickly?

In this work, we use a modified and improved version of the blockstructured adaptive mesh refinement (AMR) code ENZO [2] to provide high spatial and temporal resolution for modeling the formation of SMBHs via direct collapse within dark matter (DM) halos at high redshifts. The radiation hydrodynamics equations are solved in …

Integrative Computational Modeling For Developing Means To Manipulate Biological Cells And For Solving Complex Engineering Problems, Yu Zhao, Guigen Zhang

Commonwealth Computational Summit

Computational modeling has become more widely used to guide the design of microfluidic devices for manipulating cells using Dielectrophoresis (DEP), and devise novel means for advancing the study of cellular science and engineering. Conventionally, cells are treated as volumeless points in the system, which allows study of the movement of groups of particles under the effect of field. However, this approach often neglects the distortion effect of particle on external field, as well as interactions among particles. Moreover, it ignores the complex inner structures of cell, which are the causes of distinctive cell behavior. To more accurately model the behavior …

Dynamic Load Balancing Based On Live Virtual Machine Migration, Manh Do, Michael Galloway

Commonwealth Computational Summit

Recently, cloud computing is a new trend emerging in computer technology with a huge demand from the clients, which leads to the consumption of a tremendous amount of energy. Load balancing is taken into account as a vital part of managing income demand, improving the cloud system's performance and reducing the energy cost. Live virtual machine migration is a technique to perform the dynamic load balancing algorithm. To optimize the cloud cluster, there are three issues to consider: First, how does the cloud cluster distribute the virtual machine (VM) requests from clients to all physical machine (PM) when each machine …

Cloud-Based Text Analytics: Harvesting, Cleaning And Analyzing Corporate Earnings Conference Calls, Michael Chuancai Zhang, Vikram Gazula, Dan Stone, Hong Xie

Commonwealth Computational Summit

Does management language cohesion in earnings conference calls matter to the capital market? As a part of the research on the above question, and taking advantage of the modern IT technologies, this project:

• harvested 115,882 earnings conference call transcripts from SeekingAlpha.com
• parsed and structured 89,988 transcripts using regular expressions in Stata
• analyzed 179,976 text files using Amazon Elastic Compute Cloud (Amazon EC2), which
• saved almost 2 years (675 days) of the project time

High-Fidelity Simulations Of Water Jet In Air Crossflow, Austin Johnston, Prashant Khare

Commonwealth Computational Summit

Objectives

• Investigate detailed physics underlying liquid jets in crossflow configurations applicable to various applications such as, gas-turbine, scramjet, and afterburner fuel injection.
• Develop models to predict the statistical behaviors of resulting droplets.

Adversarial Discriminative Domain Adaptation For Extracting Protein-Protein Interactions From Text, Anthony Rios, Ramakanth Kavuluru, Zhiyong Lu

Commonwealth Computational Summit

Relation extraction is the process of extracting structured information from unstructured text. Recently, neural networks (NNs) have produced state-of-art results in extracting protein-protein interactions (PPIs) from text. While multiple corpora have been created to extract PPIs from text, most methods have shown poor cross-corpora generalization. In other words, models trained on one dataset perform poorly on other datasets for the same task. In the case of PPI, the F1 has been shown to vary by as much as 30% between different datasets. In this work, we utilize adversarial discriminative domain adaptation (ADDA) to improve the generalization between the source and …

Imputing Trust Network Information In Nmf‐Based Recommendation Systems, Fatemah Alghamedy, Jun Zhang

Commonwealth Computational Summit

With the emergence of E‐commerce, recommendation system becomes a significant tool which can help both sellers and buyers. It helps sellers by increasing the profits and advertising items to customers. In addition, recommendation systems facilitate buyers to find items they are looking for easily.

In recommendation systems, the rating matrix R represents users' ratings for items. The rows in the rating matrix represent the users and the columns represent items. If particular user rates a particular item, then the value of the intersection of the user row and item column holds the rating value. The trust matrix T describes the …

Discovery Of Sex-Specific Regions In A Salamander Genome, Nataliya Y. Timoshevskaya, Melissa C. Keinath, Jeramiah J. Smith

Commonwealth Computational Summit

Biological Aspects:

Salamander (Ambystoma mexicanum) has a gigantic genome: ~32,000,000,000 bases (10X of size of human genome)

Sex is determined by a pair of morphologically identical chromosomes:

• ZZ in male
• ZW in female

Object:

• Find (if there are any) genomic differences between chromosomes W and Z

Workflow:

1. Sequencing and de novo assembly of the reference salamander genome
2. Alignment of short sequences from male and female genomes to the reference
3. Coverage analysis

Registration And Grouping Algorithms In Protein Nmr Derived Peak Lists And Their Application In Protein Nmr Reference Correction, Andrey Smelter, Xi Chen, Eric C. Rouchka, Hunter N. B. Moseley

Commonwealth Computational Summit

Nuclear magnetic resonance spectroscopy of proteins (protein NMR) is a powerful analytical technique for studying structure and dynamics of proteins. Almost all aspects of protein NMR have been accelerated by the development of software tools that enable the analysis of NMR spectral data and its utilization in studying protein structure and dynamics. This includes software for raw NMR processing, spectral visualization, protein resonance assignment, and structure determination. However, full automation of protein NMR data analysis is still a work in progress and data analysis still requires an expert NMR spectroscopist utilizing an array of software tools.

While manual resonance assignment …

Evolution Of Barred Galaxies In Spinning Dark Matter Halos: High Resolution N-Body Simulations At Dlx, Angela Collier, Isaac Shlosman, Clayton Heller

Commonwealth Computational Summit

Observations show that galaxies are dominated by stellar disks immersed in much more massive, slowly tumbling dark matter (DM) halos. Large fraction of galactic disks, at least 75%, are barred (see Hubble Fork on the right). Stellar bars form either via spontaneous break of axial symmetry or via galaxy interactions.

The formation and evolution of stellar bars is not fully understood. Stellar bar evolution is highly nonlinear and cannot be treated analytically. The main approach to study these disk-halo systems is via numerical simulations, whose goal is to explain why galaxies have such a wide range of morphologies as shown …

Dark Matter Halo Mass Function From Hpc N-Body Simulations, Da Bi, Isaac Shlosman, Emilio Romano-Diaz

Commonwealth Computational Summit

Dark matter (DM) dominates the matter in the Universe. Because of self-gravity, DM collapses and becomes clumpy, building the large-scale hierarchical structures. Baryons assemble within DM potential wells and form galaxies.

Because we can not directly observe DM halos, numerical simulations is the only way one can study their dynamics and other properties. Using N-body simulations, we can obtain the Halo Mass Function (HMF), which provides the abundance of DM halos as a function of their mass. The HMF depends weakly on cosmological redshift and is one of the basic tools in modern cosmology.

We use GIZMO --- a flexible, …

The Lamprey Genome: Illuminating Genomic Change Across Eons And Embryogenesis, Jeramiah J. Smith, Courtney K. M. Waterbury, Melissa C. Keinath, Cody B. Saraceno, Vladimir A. Timoshevskiy, Nataliya Y. Timoshevskaya

Commonwealth Computational Summit

The lamprey genome provides unique insights into both the deep evolutionary history of vertebrate genomes and the maintenance of genome structure/integrity over development. The lamprey lineage diverged from all other vertebrates approximately 500 million years ago. As such, comparisons between lamprey and other vertebrates permit reconstruction of ancient duplication and rearrangement events that defined the fundamental architecture and gene content of all extant vertebrate genomes. Lamprey also undergoes programmatic changes genome structure that result in the physical elimination of ~20% of its genomic DNA (~0.5Gb from a ~2 Gb genome) from all somatic cell lineages during early embryonic development. Here, …

Utilizing Nvidia Gpus For Waveform Analysis For The Nab Experiment, David G. Mathews, A. P. Sprow, C. B. Crawford

Commonwealth Computational Summit

GPUs are composed of a large number of small computational cores compared to CPUs which are generally just a few larger cores. While the CPU excels at linear processes, the GPU excels at parallel tasks. For this project, the goal was to find a way to use the massive parallelism of a GPU to rapidly analyze waveform data from the Nab experiment.

How Low Can You Go? Feature Selection For Drug Discovery, Derek Jones, Sally R. Ellingson, W. A. De Jong

Commonwealth Computational Summit

The cost of bringing a drug to market depends on how quickly a candidate drug can be “discovered” and evaluated to ensure safety and effectiveness. In this work we develop a method for predicting whether a given drug and protein compound will “bind.” Our aim is to select a set of features to predict drug-protein interactions.

This study focuses on kinases. Kinase inhibitors are the largest class of new cancer therapies. Selective inhibition is difficult due to high sequence similarity, leading to off-target interactions and side-effects. Pictured here human c-SRC.

Investigation Of Photo-Responsive Titania Surfaces, Evan Hyden, Matthew J. Beck

Commonwealth Computational Summit

Titania (TiO₂) is an oxide commonly used for photocatalysis. This proclivity to form free radicals when exposed to light has also prompted interest in applying porous TiO₂ as a reactive/responsive membrane. Surfaces of both main polymorphs of TiO₂ (especially rutile {110} and anatase {101}) have been shown to be stable and photo-active.

Polyacrylic acid (PAA) is a molecule of interest as changes in pH can cause the molecule to contract or expand, due to changes in the charge distribution relating to the carboxyl group. Due to TiO₂’s ability to generate photocarriers, we hypothesize that …

Structure-Based Drug Discovery: Computational Virtual Screening, Robert C. Monsen, Lynn Deleeuw, Jon Maguire, William L. Dean, Robert D. Gray, Jonathan B. Chaires, John O. Trent

Commonwealth Computational Summit

No abstract provided.

Aberrant Coordination Geometries Discovered In Most Abundant Metalloproteins, Sen Yao, Robert M. Flight, Eric C. Rouchka, Hunter N. B. Moseley

Commonwealth Computational Summit

Metalloproteins play crucial biochemical roles in our body and are essential across all domains of life. The structural environment around a metal ion, especially the coordination geometry (CG), is both sequentially and functionally relevant. Studies of the metalloprotein’s CG will greatly help alleviate the imbalance between the ample sequence data available and the insufficient knowledge on protein functions. Current methodologies in characterizing metalloproteins’ CG consider only previously reported CG (canonical CG) models based primarily on nonbiological chemical context. Exceptions to these canonical CG models can greatly hamper the ability to characterize metalloproteins both structurally and functionally.

Use Of Hpc To Analyze Changes In Gene Expression During Fruit Fly Spermiogenesis, Sepideh Dadkhah, Douglas A. Harrison, Jeramiah J. Smith

Commonwealth Computational Summit

In the fruit fly, Drosophila melanogaster, JAK/STAT signaling during spermiogenesis is known to play a crucial role in the maintenance of stem cells of the testis. Recent studies in our lab have shown that activation of the JAK/STAT pathway in somatic cyst cells is also required for the later stages of spermiogenesis like individualization.

The main goal of this project is to characterize the events downstream of JAK/STAT signaling in spermiogenesis and more specifically to determine the mechanism by which JAK/STAT activation regulates individualization, a later stage in spermiogenesis where 64 individual spermatids are formed from a 64-interconnected spermatid …

Spectroscopic And Computational Characterization Of Lanthanum-Mediated C-H And N-H Bond Activation Of Amines, Silver Nyambo, Yuchen Zhang, Dong-Sheng Yang

Commonwealth Computational Summit

Metal-mediated bond activation of small organic and inorganic molecules plays critical roles in chemical transformation of small molecules into value-added products.

This is because few of such chemical reactions would occur under mild conditions without the metal activation.

In this work, lanthanum atom reactions with alkylamines are carried out in a laser-ablation supersonic molecular beam source; C-H and N-H bond activation in these species is investigated.

The reaction products are observed with photoionization time-of-flight mass spectrometry and characterized by mass-analyzed threshold ionization (MATI) spectroscopy and theoretical calculations.

Adiabatic ionization energy and metal-ligand and ligand-based vibrational frequencies of several short-lived lanthanum …

Evolution Of Bias In Human And Machine Learning Algorithm Interaction, Wenlong Sun, Olfa Nasraoui, Patrick Shafto

Commonwealth Computational Summit

Human algorithm interaction:

• people are now affected by the output of all types of machine learning algorithms.
• social media, blogs, social networks, and other services and applications.

Motivation:

• ML algorithm relied on reliable labels from experts to build prediction.
• However, ML algorithm started to receive data from the more general population.
• The interaction leads to biased result which is caused by ingesting unchecked information from general population, such as biased samples and biased labels.

Using Explainability For Constrained Matrix Factorization, Behnoush Abdollahi, Olfa Nasraoui

Commonwealth Computational Summit

Explainable Model

• Black Box (opaque) predictors such as Deep learning and Matrix Factorization are accurate,
• ... but lack interpretability and ability to give explanations.
• White Box models such as rules and decision trees are interpretable (explainable),
• ... but lack accuracy.

Distributed Object Segmentation In Big Spatial Data, Hamid Hamraz

Commonwealth Computational Summit

Problem Overview

• Big spatial data is delivered in the shape of several tiles and requires distributed computing.
• Object segmentation is not perfectly parallel because the objects may lie across the tiles.

Ontology Modeling 2.0: Next Steps, Pascal Hitzler

Commonwealth Computational Summit

Semantic Web as a field of research and applications is concerned with methods and tools for data sharing, discovery, integration, and reuse, both on and off the World Wide Web. In the form of knowledge graphs and their underlying schemas, Semantic Web technologies are currently entering industrial mainstream. At the same time, the ever increasing prevalence of publicly available structured data on the Semantic Web enables new applications in a variety of domains, and as part of this presentation, we provide a conceptual approach that leverages such data in order to explain the input-output behavior of trained artificial neural networks. …

Welcome Morning Talk, Anthony Elam, Lisa A. Cassis

Commonwealth Computational Summit

No abstract provided.

Differentiating The Impact Of Nitrogen Chemical States On Optical Properties Of Nitrogen-Doped Graphene Quantum Dots, Timothy Pillar-Little, Doo Young Kim

Chemistry Faculty Publications

The optical properties of top-down synthesized oxidized graphene quantum dots (ox-GQDs) and nitrogen-incorporating graphene quantum dots (N-GQDs) along a range of hydrothermal treatment temperatures were observed. By controlling the hydrothermal treatment temperature, different chemical states of nitrogen atoms were incorporated into GQDs. Below 150 °C, edge-terminating amines and amides dominated the nitrogen content of N-GQDs. Above 150 °C, nitrogen was primarily present in the forms of pyridinic, pyrrolic and quaternary N. In addition to the absorbance and emission profiles of ox-GQDs and N-GQDs, pH-dependent emission spectra were collected to probe chemical states of nitrogen atoms and investigate the relationship between …

Bulk Properties Of The Medium Produced In Relativistic Heavy-Ion Collisions From The Beam Energy Scan Program, L. Adamczyk, James K. Adkins, G. Agakishiev, M. M. Aggarwal, Z. Ahammed, N. N. Ajitanand, I. Alekseev, D. M. Anderson, R. Aoyama, A. Aparin, D. Arkhipkin, E. C. Aschenauer, M. U. Ashraf, A. Attri, G. S. Averichev, X. Bai, V. Bairathi, A. Behera, R. Bellwied, A. Bhasin, A. K. Bhati, P. Bhattarai, J. Bielcik, J. Bielcikova, L. C. Bland, I. G. Bordyuzhin, J. Bouchet, J. D. Brandenburg, A. V. Brandin, D. Brown, Renee H. Fatemi, Suvarna Ramachandran

Physics and Astronomy Faculty Publications

We present measurements of bulk properties of the matter produced in Au+Au collisions at √S_NN = 7.7, 11.5, 19.6, 27, and 39 GeV using identified hadrons (π^±, K^±, p, and p¯) from the STAR experiment in the Beam Energy Scan (BES) Program at the Relativistic Heavy Ion Collider (RHIC). Midrapidity (|y| < 0.1) results for multiplicity densities dN/dy, average transverse momenta ⟨p_T⟩, and particle ratios are presented. The chemical and kinetic freeze-out dynamics at these energies are discussed and presented as a function of collision centrality and energy. These results constitute the systematic …

Screening And Plasma Oscillations In An Electron Gas In The Hydrodynamic Approximation, Eugene B. Kolomeisky, Joseph P. Straley

Physics and Astronomy Faculty Publications

A hydrodynamic theory of screening in a generic electron gas of arbitrary dimensionality is given that encompasses all previously studied cases and clarifies the predictions of the many-body approach. We find that long-wavelength plasma oscillations are classical phenomena with quantum-mechanical effects playing no explicit role. The character of the oscillations is solely dictated by the dimensionality of the electron system and its equation of state in the neutral limit. Materials whose excitations are described by the Dirac dispersion law—such as doped graphene or a Weyl semimetal—are no exception to this rule.