arXiv subjects

Caleb Phillips

Publications and source records attributed to Caleb Phillips.

8 recordsLinked to original sources

arXiv2021

Research Data Infrastructure for High-Throughput Experimental Materials Science

The High-Throughput Experimental Materials Database (HTEM-DB) is the endpoint repository for inorganic thin-film materials data collected during combinatorial experiments at the National Renewable Energy Laboratory (NREL). This unique data asset is enabled by the Research Data Infrastructure (RDI) - a set of custom data tools that collect, process, and store experimental data and metadata. Here, we describe the experimental data-tool workflow from the RDI to the HTEM-DB to illustrate the strategies and best practices currently used for materials data at NREL. Integration of these data tools with the experimental processes establishes a data communication pipeline between experimental and data science communities. In doing so, this work motivates the creation of similar data workflows at other institutions to aggregate valuable data and increase its usefulness for future data studies. These types of investments can greatly accelerate the pace of learning and discovery in the materials science field, by making data accessible to new and rapidly evolving data methods.

cond-mat.mtrl-sci↗

arXiv2019

VASTA: A Vision and Language-assisted Smartphone Task Automation System

We present VASTA, a novel vision and language-assisted Programming By Demonstration (PBD) system for smartphone task automation. Development of a robust PBD automation system requires overcoming three key challenges: first, how to make a particular demonstration robust to positional and visual changes in the user interface (UI) elements; secondly, how to recognize changes in the automation parameters to make the demonstration as generalizable as possible; and thirdly, how to recognize from the user utterance what automation the user wishes to carry out. To address the first challenge, VASTA leverages state-of-the-art computer vision techniques, including object detection and optical character recognition, to accurately label interactions demonstrated by a user, without relying on the underlying UI structures. To address the second and third challenges, VASTA takes advantage of advanced natural language understanding algorithms for analyzing the user utterance to trigger the VASTA automation scripts, and to determine the automation parameters for generalization. We run an initial user study that demonstrates the effectiveness of VASTA at clustering user utterances, understanding changes in the automation parameters, detecting desired UI elements, and, most importantly, automating various tasks. A demo video of the system is available here: http://y2u.be/kr2xE-FixjI

cs.HC↗

arXiv2019

Wrist02 -- Reliable Peripheral Oxygen Saturation Readings from Wrist-Worn Pulse Oximeters

Peripheral blood oxygen saturation Sp02 is a vital measure in healthcare. Modern off-the-shelf wrist-worn devices, such as the Apple Watch, FitBit, and Samsung Gear, have an onboard sensor called a pulse oximeter. While pulse oximeters are capable of measuring both Sp02 and heart rate, current wrist-worn devices use them only to determine heart rate, as Sp02 measurements collected from the wrist are believed to be inaccurate. Enabling oxygen saturation monitoring on wearable devices would make these devices tremendously more useful for health monitoring and open up new avenues of research. To the best of our knowledge, we present the first study of the reliability of Sp02 sensing from the wrist. Using a custom-built wrist-worn pulse oximeter, we find that existing algorithms designed for fingertip sensing are a poor match for this setting, and can lead to over 90% of readings being inaccurate and unusable. We further show that sensor placement and skin tone have a substantial effect on the measurement error, and must be considered when designing wrist-worn Sp02 sensors and measurement algorithms. Based on our findings, we propose \codename, an alternative approach for reliable Sp02 sensing. By selectively pruning data, \codename achieves an order of magnitude reduction in error compared to existing algorithms, while still providing sufficiently frequent readings for continuous health monitoring.

cs.CY↗

arXiv2018

In Vitro Vascularized Tumor Platform for Modeling Tumor-Vasculature Interactions of Inflammatory Breast Cancer

Inflammatory breast cancer (IBC), a rare form of breast cancer associated with increased angiogenesis and metastasis, is largely driven by tumor-stromal interactions with the vasculature and the extracellular matrix (ECM). However, there is currently a lack of understanding of the role these interactions play in initiation and progression of the disease. In this study, we developed the first three-dimensional, in vitro, vascularized, breast tumor platform to quantify the spatial and temporal dynamics of tumor-vasculature and tumor-ECM interactions specific to IBC. Platforms consisting of collagen type 1 ECM with an endothelialized blood vessel were cultured with IBC cells, MDA-IBC3 (HER2+) or SUM149 (triple negative), and for comparison to non-IBC cells, MDA-MB-231 (triple negative). An acellular collagen platform with an endothelial blood vessel served as control. SUM149 and MDA-MB-231 platforms exhibited a significantly (p<0.05) higher vessel permeability and decreased endothelial coverage of the vessel lumen compared to the control. Both IBC platforms, MDA-IBC3 and SUM149, expressed higher levels of VEGF (p<0.05) and increased collagen ECM porosity compared to non-IBC MDA-MB-231 (p<0.05) and control (p<0.01) platforms. Additionally, unique to the MDA-IBC3 platform, we observed progressive sprouting of the endothelium over time resulting in viable vessels with lumen. The newly sprouted vessels encircled clusters of MDA-IBC3 cells replicating a feature of in vivo IBC. The IBC in vitro vascularized platforms introduced in this study model well-described in vivo and clinical IBC phenotypes and provide an adaptable, high throughout tool for systematically and quantitatively investigating tumor-stromal mechanisms and dynamics of tumor progression.

q-bio.CB↗

arXiv2018

Message-passing neural networks for high-throughput polymer screening

Machine learning methods have shown promise in predicting molecular properties, and given sufficient training data machine learning approaches can enable rapid high-throughput virtual screening of large libraries of compounds. Graph-based neural network architectures have emerged in recent years as the most successful approach for predictions based on molecular structure, and have consistently achieved the best performance on benchmark quantum chemical datasets. However, these models have typically required optimized 3D structural information for the molecule to achieve the highest accuracy. These 3D geometries are costly to compute for high levels of theory, limiting the applicability and practicality of machine learning methods in high-throughput screening applications. In this study, we present a new database of candidate molecules for organic photovoltaic applications, comprising approximately 91,000 unique chemical structures.Compared to existing datasets, this dataset contains substantially larger molecules (up to 200 atoms) as well as extrapolated properties for long polymer chains. We show that message-passing neural networks trained with and without 3D structural information for these molecules achieve similar accuracy, comparable to state-of-the-art methods on existing benchmark datasets. These results therefore emphasize that for larger molecules with practical applications, near-optimal prediction results can be obtained without using optimized 3D geometry as an input. We further show that learned molecular representations can be leveraged to reduce the training data required to transfer predictions to a new DFT functional.

physics.comp-ph↗

arXiv2011

Strange Beta: An Assistance System for Indoor Rock Climbing Route Setting Using Chaotic Variations and Machine Learning

This paper applies machine learning and the mathematics of chaos to the task of designing indoor rock-climbing routes. Chaotic variation has been used to great advantage on music and dance, but the challenges here are quite different, beginning with the representation. We present a formalized system for transcribing rock climbing problems, then describe a variation generator that is designed to support human route-setters in designing new and interesting climbing problems. This variation generator, termed Strange Beta, combines chaos and machine learning, using the former to introduce novelty and the latter to smooth transitions in a manner that is consistent with the style of the climbs This entails parsing the domain-specific natural language that rock climbers use to describe routes and movement and then learning the patterns in the results. We validated this approach with a pilot study in a small university rock climbing gym, followed by a large blinded study in a commercial climbing gym, in cooperation with experienced climbers and expert route setters. The results show that {\sc Strange Beta} can help a human setter produce routes that are at least as good as, and in some cases better than, those produced in the traditional manner.

cs.AI↗

arXiv2011

The Stability of The Longley-Rice Irregular Terrain Model for Typical Problems

In this paper, we analyze the numerical stability of the popular Longley-Rice Irregular Terrain Model (ITM). This model is widely used to plan wireless networks and in simulation-validated research and hence its stability is of fundamental importance to the correctness of a large amount of work. We take a systematic approach by first porting the reference ITM implementation to a multiprecision framework and then generating loss predictions along many random paths using real terrain data. We find that the ITM is not unstable for common numerical precisions and practical prediction scenarios.

cs.NI↗

arXiv2011

Food Redistribution as Optimization

In this paper we study the simultaneous problems of food waste and hunger in the context of the possible solution of food (waste) rescue and redistribution. To this end, we develop an empirical model that can be used in Monte Carlo simulations to study the dynamics of the underlying problem. Our model's parameters are derived from a unique data set provided by a large food bank and food rescue organization in north central Colorado. We find that food supply is a non-parametric heavy-tailed process that is well-modeled with an extreme value peaks-over-threshold model. Although the underlying process is stochastic, the basic approach of food rescue and redistribution appears to be feasible both at small and large scales. The ultimate efficacy of this model is intimately tied to the rate at which food expires and hence the ability to preserve and quickly transport and redistribute food. The cost of the redistribution is tied to the number and density of participating suppliers, and costs can be reduced (and supply increased) simply by recruiting additional donors to participate. Our results show that with sufficient funding and manpower, a significant amount of food can be rescued from the waste stream and used to feed the hungry.

cs.OH↗