Demeter
Demeter is a fully extensible, community-based land use land cover (LULC) downscaling and change detection model that allows for customization and/or additions from its user base. The model was created to generate high-resolution gridded time series representations of LULC projections from global, national, and regional models. The downscaling is based on a number of user-defined rules and drivers, including transitions priorities (i.e., crop expansion preferentially into grasslands rather than forests) and spatial constraints (e.g., nutrient availability). Demeter downscales land allocation projections for a region and/or zone to the resolution of the observed spatial data (MODIS, NLCD, etc.) using an intensification and expansion process applied to each LULC type. Land classes from both the observed and projected data are harmonized by a user-defined binning of each class into the desired output LULC types. Continued model development is underway that will accommodate expanded output formats to enable coupling with the Soil and Water Assessment Tool (SWAT), the Community Land Model (CLM), and others. Integration with the Community Surface Dynamics Modeling System (CSDMS) is also underway by wrapping Demeter in a Basic Model Interface (BMI) to allow for component-based coupling.
nmrfit
Our Python package, nmrfit, reads the output from an nuclear magnetic resonance (NMR) experiment and, through a number of intuitive API calls, produces a least-squares fit of Voigt-function approximations via particle swarm optimization. Fitted peaks can then be used to perform quantitative NMR analysis, including isotope ratio approximation. There are two alternatives in this space (nmrglue and lmfit), but both are insufficient in that: Voigt functions are not area parameterized, limiting signal contribution quantification accuracy (necessary to calculate isotope ratios). Optimization methods are not robust enough for isotope ratio applications. For example, carbon-12 peaks dominate carbon-13 peaks in terms of relative intensity, causing many optimization methods to fit the former extremely well, but largely ignore the latter. Our software includes a dynamic-weighting scheme and optimization by particle swarm to robustly and faithfully fit peaks regardless of relative intensity differences. Initial parameter selection tools are either lacking or insufficient. Our software supplies utilities for bounds selection, automatic or manual peak selection, and a number of options for initial phase correction. In-phase data is required. Our approach requires an initial, approximate phase correction for seed parameter determination, but the zeroth- and first- order phase are fit parameters, which is novel in this space. Supplied APIs are not particularly intuitive. Our offering requires only ~10 lines of code to go from data input to publication-quality plots, all through intuitive function and method calls.
ErrFilter
ErrFilter is a tool that allows an analyst to explore data by defining simple Boolean filters on data. The tool also allows the user to sort the results by any attribute. The user sees both the distribution of filtered data as well as selected examples.
A hierarchical clustering scheme for atom tomography data
The current scripts utilize a well-known cluster analysis algorithm, called ordering points to identify the clustering structures (OPTICS), and an automatic cluster extraction algorithm for clustering analysis of 3D spatial data, specifically for the Atom Probe Tomography data. A hierarchical clustering scheme, together with density-based clustering, were implemented. The advantages of the developed algorithm is that the produced a hierarchy of clusters, which can be retrieved and re-analyzed later, could save researchers time in manually exploring parameter space as in current widely used method, and the capability to identify clusters of varying atomic density in a single run. The current developed scripts can also serve as a platform to test new algorithms with the internal and external APT experts.
DefectSegNet
Microsoft Word - SREP-19-11906_manuscript_revised.doc We demonstrate the feasibility of automated identification of common crystallographic defects in steels using deep learning semantic segmentation, based on high-quality microscopy data. In particular, the DefectSegNet - a new hybrid CNN architecture with skip connections within and across the encoder and decoder was developed, and has proved to be effective at perceptual defect identification with high pixel-wise accuracy.
UnityMol-APBS (NIH iEdison No. 0685901-19-0009, Grant No. GM069702)
Virtual reality is a powerful tool with the ability to immerse a user within a completely external environment. This immersion is particularly useful when visualizing and analyzing interactions between small organic molecules, molecular inorganic complexes, and biomolecular systems such as redox proteins and enzymes. A common tool used in the biomedical community to analyze such interactions is the Adaptive Poisson-Boltzmann Solver (APBS) software, which was developed to solve the equations of continuum electrostatics for large biomolecular assemblages. Numerous applications exist for using APBS in the biomedical community including analysis of protein ligand interactions and APBS has enjoyed widespread adoption throughout this biomedical community. Currently, typical use of the full APBS toolset is completed via the command line followed by visualization using a variety of twodimensional external molecular visualization software. This process has inherent limitations: visualization of three-dimensional objects using a twodimensional interface masks important information within the depth component. Herein, we have developed a single application, UnityMol-APBS, that provides a dual experience where users can utilize the full range of the APBS toolset, without the use of a command line interface, by use of a simple graphical user interface (GUI) for either a standard desktop or immersive virtual reality experience.
NWPEsSe (North West Potential Energy Surface Search Engine)
Global optimization of nanosized clusters is an important and fundamental problem in theoretical studies in many chemical fields, like catalysis, material, or energy chemistry, etc. In this paper, the powerful artificial bee colony (ABC) algorithm, which has been applied successfully in the global optimization of atomic and molecular clusters, has been developed for nanosized clusters of complex structures. The new ABC algorithm is applied to the global optimization of 4 systems of different chemical nature: gas phase Au55, ligated Au82+, graphene oxide and defected rutile-supported Au8, and cluster assemble . These clusters have sizes lie between 1 to 3 nm and contain up to 1000 atoms, raising great challenges to the algorithm. Reliable global minima (GMs) are obtained for all cases, some of which are better than those reported in literature, indicating the excellent performance of the new ABC algorithm. These GMs provide chemically important insights into the systems. The new ABC algorithm has been coded into the latest version of ABCluster, making it a promise tool for chemists from broad fields to rapidly carry out global optimizations of nanosized clusters.
Mcqueuer (mcqr) - Open Source
Mcqueuer is a simple tool that allows anyone from researchers to experienced developers to create multi-node/multi-core jobs by simply creating a file with a list of commands. Users simply combine tasks, which would otherwise each be their own job on the cluster, into a single file that is given to Mcqueuer. Mcqueuer then does the heavy lifting required to process the tasks in parallel in a single multi-node job. In addition, Mcqueuer provides load-balancing, which frees the user from having to worry about complex memory and CPU considerations, and instead focus on the processing itself.
NOUS: A Knowledge Graph Management System - Open Source Copuright
We developed a system for automated construction of knowledge graphs from text data. Specifically, the following are unique features of this system. 1. Feature extraction from raw text and clustering based reduction of synonymous, redundant facts. 2. Link prediction based approach towards confidence scoring of extracted facts. 3. Large-scale profiling of entities in the knowledge graph. 4. Generating natural language based explanation to a restricted class of questions.