arXiv subjects

Shashi Shekhar

Publications and source records attributed to Shashi Shekhar.

33 records · Page 2Linked to original sources

arXiv2022

SAMCNet for Spatial-configuration-based Classification: A Summary of Results

The goal of spatial-configuration-based classification is to build a classifier to distinguish two classes (e.g., responder, non-responder) based on the spatial arrangements (e.g., spatial interactions between different point categories) given multi-category point data from two classes. This problem is important for generating hypotheses in medical pathology towards discovering new immunotherapies for cancer treatment as well as for other applications in biomedical research and microbial ecology. This problem is challenging due to an exponential number of category subsets which may vary in the strength of spatial interactions. Most prior efforts on using human selected spatial association measures may not be sufficient for capturing the relevant (e.g., surrounded by) spatial interactions which may be of biological significance. In addition, the related deep neural networks are limited to category pairs and do not explore larger subsets of point categories. To overcome these limitations, we propose a Spatial-interaction Aware Multi-Category deep neural Network (SAMCNet) architecture and contribute novel local reference frame characterization and point pair prioritization layers for spatial-configuration-based classification. Extensive experimental results on multiple cancer datasets show that the proposed architecture provides higher prediction accuracy over baseline methods.

cs.CV↗

arXiv2022

Understanding COVID-19 Effects on Mobility: A Community-Engaged Approach

Given aggregated mobile device data, the goal is to understand the impact of COVID-19 policy interventions on mobility. This problem is vital due to important societal use cases, such as safely reopening the economy. Challenges include understanding and interpreting questions of interest to policymakers, cross-jurisdictional variability in choice and time of interventions, the large data volume, and unknown sampling bias. The related work has explored the COVID-19 impact on travel distance, time spent at home, and the number of visitors at different points of interest. However, many policymakers are interested in long-duration visits to high-risk business categories and understanding the spatial selection bias to interpret summary reports. We provide an Entity Relationship diagram, system architecture, and implementation to support queries on long-duration visits in addition to fine resolution device count maps to understand spatial bias. We closely collaborated with policymakers to derive the system requirements and evaluate the system components, the summary reports, and visualizations.

cs.CY↗

arXiv2021

Towards Comparative Physical Interpretation of Spatial Variability Aware Neural Networks: A Summary of Results

Given Spatial Variability Aware Neural Networks (SVANNs), the goal is to investigate mathematical (or computational) models for comparative physical interpretation towards their transparency (e.g., simulatibility, decomposability and algorithmic transparency). This problem is important due to important use-cases such as reusability, debugging, and explainability to a jury in a court of law. Challenges include a large number of model parameters, vacuous bounds on generalization performance of neural networks, risk of overfitting, sensitivity to noise, etc., which all detract from the ability to interpret the models. Related work on either model-specific or model-agnostic post-hoc interpretation is limited due to a lack of consideration of physical constraints (e.g., mass balance) and properties (e.g., second law of geography). This work investigates physical interpretation of SVANNs using novel comparative approaches based on geographically heterogeneous features. The proposed approach on feature-based physical interpretation is evaluated using a case-study on wetland mapping. The proposed physical interpretation improves the transparency of SVANN models and the analytical results highlight the trade-off between model transparency and model performance (e.g., F1-score). We also describe an interpretation based on geographically heterogeneous processes modeled as partial differential equations (PDEs).

cs.LG↗

arXiv2021

Statistically-Robust Clustering Techniques for Mapping Spatial Hotspots: A Survey

Mapping of spatial hotspots, i.e., regions with significantly higher rates of generating cases of certain events (e.g., disease or crime cases), is an important task in diverse societal domains, including public health, public safety, transportation, agriculture, environmental science, etc. Clustering techniques required by these domains differ from traditional clustering methods due to the high economic and social costs of spurious results (e.g., false alarms of crime clusters). As a result, statistical rigor is needed explicitly to control the rate of spurious detections. To address this challenge, techniques for statistically-robust clustering (e.g., scan statistics) have been extensively studied by the data mining and statistics communities. In this survey we present an up-to-date and detailed review of the models and algorithms developed by this field. We first present a general taxonomy for statistically-robust clustering, covering key steps of data and statistical modeling, region enumeration and maximization, and significance testing. We further discuss different paradigms and methods within each of the key steps. Finally, we highlight research gaps and potential future directions, which may serve as a stepping stone in generating new ideas and thoughts in this growing field and beyond.

stat.ML↗

arXiv2021

Vehicle Emissions Prediction with Physics-Aware AI Models: Preliminary Results

Given an on-board diagnostics (OBD) dataset and a physics-based emissions prediction model, this paper aims to develop an accurate and computational-efficient AI (Artificial Intelligence) method that predicts vehicle emissions. The problem is of societal importance because vehicular emissions lead to climate change and impact human health. This problem is challenging because the OBD data does not contain enough parameters needed by high-order physics models. Conversely, related work has shown that low-order physics models have poor predictive accuracy when using available OBD data. This paper uses a divergent window co-occurrence pattern detection method to develop a spatiotemporal variability-aware AI model for predicting emission values from the OBD datasets. We conducted a case study using real-world OBD data from a local public transportation agency. Results show that the proposed AI method has approximately 65% improved predictive accuracy than a non-AI low-order physics model and is approximately 35% more accurate than a baseline model.

cs.AI↗

arXiv2021

A National Research Agenda for Intelligent Infrastructure: 2021 Update

Strategic, sustained Federal investments in intelligent infrastructure will increase safety and resilience, improve efficiencies and civic services, and broaden employment opportunities and job growth nationwide. The technologies that comprise intelligent infrastructure can also provide keys to solving some of the most vexing challenges we face today, including confronting future pandemics and natural disasters, achieving sustainability and energy efficiency goals, and advancing social justice. Enabling those technologies effectively will require investment in the associated computing research as well, beyond and in concert with the basic building projects. In 2017, the Computing Community Consortium (CCC) produced a series of intelligent infrastructure whitepapers, and in 2020 CCC issued a set of companion whitepapers on closely related topics. Here we briefly survey those earlier works, and then highlight four themes of rising national prominence where intelligent infrastructure can also play an enabling role, driven by experiences with the COVID-19 pandemic and the social justice movement. We conclude with recommendations for the necessary research investments.

cs.CY↗

arXiv2020

Spatially correlated rotational dynamics reveals strain dependence in amorphous particle packings

Microstructural dynamics in amorphous particle packings is commonly probed by quantifying particle displacements. While rigidity in particle packings emerges when displacement of particles are hindered, it is not obvious how the typically disordered displacement metrics connect to mechanical response. Particle rotations, in contrast, are much less sensitive to confinement effects, while still sensitive to the mechanics of the packing. So far, little attention has been paid to connect microscopic rotational motion to mechanics of athermal amorphous packings. We demonstrate through experimental measurements that particle packing mechanics can be directly linked to the rotational motion of even round particles in a sheared packing. Our results show that the diffusive nature of rotational dynamics is highly strain sensitive. Additionally, there is substantial spatial correlation in rotation dynamics that is a function of the particle friction and packing density. Analysis of our measurements reveals that particle rotation dynamics plays an essential role in amorphous material mechanics.

cond-mat.soft↗

arXiv2020

Towards Spatial Variability Aware Deep Neural Networks (SVANN): A Summary of Results

Spatial variability has been observed in many geo-phenomena including climatic zones, USDA plant hardiness zones, and terrestrial habitat types (e.g., forest, grasslands, wetlands, and deserts). However, current deep learning methods follow a spatial-one-size-fits-all(OSFA) approach to train single deep neural network models that do not account for spatial variability. In this work, we propose and investigate a spatial-variability aware deep neural network(SVANN) approach, where distinct deep neural network models are built for each geographic area. We evaluate this approach using aerial imagery from two geographic areas for the task of mapping urban gardens. The experimental results show that SVANN provides better performance than OSFA in terms of precision, recall,and F1-score to identify urban gardens.

cs.CV↗

arXiv2020

A Physics Model-Guided Online Bayesian Framework for Energy Management of Extended Range Electric Delivery Vehicles

Increasing the fuel economy of hybrid electric vehicles (HEVs) and extended range electric vehicles (EREVs) through optimization-based energy management strategies (EMS) has been an active research area in transportation. However, it is difficult to apply optimization-based EMS to current in-use EREVs because insufficient knowledge is known about future trips, and because such methods are computationally expensive for large-scale deployment. As a result, most past research has been validated on standard driving cycles or on recorded high-resolution data from past real driving cycles. This paper improves an in-use rule-based EMS that is used in a delivery vehicle fleet equipped with two-way vehicle-to-cloud connectivity. A physics model-guided online Bayesian framework is described and validated on large number of in-use driving samples of EREVs used for last-mile package delivery. The framework includes: a database, a preprocessing module, a vehicle model and an online Bayesian algorithm module. It uses historical 0.2 Hz resolution trip data as input and outputs an updated parameter to the engine control logic on the vehicle to reduce fuel consumption on the next trip. The key contribution of this work is a framework that provides an immediate solution for fuel use reduction of in-use EREVs. The framework was also demonstrated on real-world EREVs delivery vehicles operating on actual routes. The results show an average of 12.8% fuel use reduction among tested vehicles for 155 real delivery trips. The presented framework is extendable to other EREV applications including passenger vehicles, transit buses, and other vocational vehicles whose trips are similar day-to-day.

eess.SP↗

arXiv2020

Adversarial Attacks on Reinforcement Learning based Energy Management Systems of Extended Range Electric Delivery Vehicles

Adversarial examples are firstly investigated in the area of computer vision: by adding some carefully designed ''noise'' to the original input image, the perturbed image that cannot be distinguished from the original one by human, can fool a well-trained classifier easily. In recent years, researchers also demonstrated that adversarial examples can mislead deep reinforcement learning (DRL) agents on playing video games using image inputs with similar methods. However, although DRL has been more and more popular in the area of intelligent transportation systems, there is little research investigating the impacts of adversarial attacks on them, especially for algorithms that do not take images as inputs. In this work, we investigated several fast methods to generate adversarial examples to significantly degrade the performance of a well-trained DRL- based energy management system of an extended range electric delivery vehicle. The perturbed inputs are low-dimensional state representations and close to the original inputs quantified by different kinds of norms. Our work shows that, to apply DRL agents on real-world transportation systems, adversarial examples in the form of cyber-attack should be considered carefully, especially for applications that may lead to serious safety issues.

cs.LG↗

arXiv2017

Theory-guided Data Science: A New Paradigm for Scientific Discovery from Data

Data science models, although successful in a number of commercial domains, have had limited applicability in scientific problems involving complex physical phenomena. Theory-guided data science (TGDS) is an emerging paradigm that aims to leverage the wealth of scientific knowledge for improving the effectiveness of data science models in enabling scientific discovery. The overarching vision of TGDS is to introduce scientific consistency as an essential component for learning generalizable models. Further, by producing scientifically interpretable models, TGDS aims to advance our scientific understanding by discovering novel domain insights. Indeed, the paradigm of TGDS has started to gain prominence in a number of scientific disciplines such as turbulence modeling, material discovery, quantum chemistry, bio-medical science, bio-marker discovery, climate science, and hydrology. In this paper, we formally conceptualize the paradigm of TGDS and present a taxonomy of research themes in TGDS. We describe several approaches for integrating domain knowledge in different research themes using illustrative examples from different disciplines. We also highlight some of the promising avenues of novel research for realizing the full potential of theory-guided data science.

cs.LG↗

arXiv2017

Intelligent Infrastructure for Smart Agriculture: An Integrated Food, Energy and Water System

Agriculture provides economic opportunity through innovation; helps rural America to thrive; promotes agricultural production that better nourishes Americans; and aims to preserve natural resources through healthy private working lands, conservation, improved watersheds, and restored forests. From agricultural production to food supply, agriculture supports rural and urban economies across the U.S. It accounts for 10% of U.S. jobs and is currently creating new jobs in the growing field of data-driven farming. However, U.S. global competitiveness associated with food and nutrition security is at risk because of accelerated investments by many other countries in agriculture, food, energy, and resource management. To ensure U.S. global competitiveness and long-term food security, it is imperative that we build sustainable physical and cyber infrastructures to enable self-managing and sustainable farming. Such infrastructures should enable next generation precision-farms by harnessing modern and emerging technologies such as small satellites, broadband Internet, tele-operation, augmented reality, advanced data analytics, sensors, and robotics.

cs.CY↗

arXiv2012

AdSplit: Separating smartphone advertising from applications

A wide variety of smartphone applications today rely on third-party advertising services, which provide libraries that are linked into the hosting application. This situation is undesirable for both the application author and the advertiser. Advertising libraries require additional permissions, resulting in additional permission requests to users. Likewise, a malicious application could simulate the behavior of the advertising library, forging the user's interaction and effectively stealing money from the advertiser. This paper describes AdSplit, where we extended Android to allow an application and its advertising to run as separate processes, under separate user-ids, eliminating the need for applications to request permissions on behalf of their advertising libraries. We also leverage mechanisms from Quire to allow the remote server to validate the authenticity of client-side behavior. In this paper, we quantify the degree of permission bloat caused by advertising, with a study of thousands of downloaded apps. AdSplit automatically recompiles apps to extract their ad services, and we measure minimal runtime overhead. We also observe that most ad libraries just embed an HTML widget within and describe how AdSplit can be designed with this in mind to avoid any need for ads to have native code.

cs.OS↗

arXiv2011

Quire: Lightweight Provenance for Smart Phone Operating Systems

Smartphone apps often run with full privileges to access the network and sensitive local resources, making it difficult for remote systems to have any trust in the provenance of network connections they receive. Even within the phone, different apps with different privileges can communicate with one another, allowing one app to trick another into improperly exercising its privileges (a Confused Deputy attack). In Quire, we engineered two new security mechanisms into Android to address these issues. First, we track the call chain of IPCs, allowing an app the choice of operating with the diminished privileges of its callers or to act explicitly on its own behalf. Second, a lightweight signature scheme allows any app to create a signed statement that can be verified anywhere inside the phone. Both of these mechanisms are reflected in network RPCs, allowing remote systems visibility into the state of the phone when an RPC is made. We demonstrate the usefulness of Quire with two example applications. We built an advertising service, running distinctly from the app which wants to display ads, which can validate clicks passed to it from its host. We also built a payment service, allowing an app to issue a request which the payment service validates with the user. An app cannot not forge a payment request by directly connecting to the remote server, nor can the local payment service tamper with the request.

cs.CR↗

arXiv2010

Minimum Spanning Tree on Spatio-Temporal Networks

Given a spatio-temporal network (ST network) where edge properties vary with time, a time-sub-interval minimum spanning tree (TSMST) is a collection of minimum spanning trees of the ST network, where each tree is associated with a time interval. During this time interval, the total cost of tree is least among all the spanning trees. The TSMST problem aims to identify a collection of distinct minimum spanning trees and their respective time-sub-intervals under the constraint that the edge weight functions are piecewise linear. This is an important problem in ST network application domains such as wireless sensor networks (e.g., energy efficient routing). Computing TSMST is challenging because the ranking of candidate spanning trees is non-stationary over a given time interval. Existing methods such as dynamic graph algorithms and kinetic data structures assume separable edge weight functions. In contrast, we propose novel algorithms to find TSMST for large ST networks by accounting for both separable and non-separable piecewise linear edge weight functions. The algorithms are based on the ordering of edges in edge-order-intervals and intersection points of edge weight functions.

cs.DS↗