Skip to main content
Smart Contract
Abstract

Implementation of Smart Contract

    In this paper, we implemented the architecture for a Smart Contract software, which works on blockchain technology. In addition, we discussed technologies used in the implementation of our system. Smart contracts have the benefit of speed, real-time updates, accuracy, lower execution risk, fewer intermediaries, lower cost, and new operational models. Since blockchain traditionally depended on mining to add transaction onto a ledger which was time consuming and required a lot of resources, we implemented an infrastructure which is mining free. Integrating blockchain technology with a traditional contract brings all the advantages of blockchain into contracts.

Reservior-Based Learning
Abstract

Echo-State and Gradient-Based Deep Learning Models for Diagnosis of Parkinson's Disease

    Reservoir Computing had been introduced in the 1990s as an alternative to trainable recurrent neural networks for sequential tasks. In recent years, due to advances in deep learning, there has been great progress on sequential tasks, in conjunction with new high performance computing architectures such as Gated Recurrent Units. In this research, we examine reservoir computing in the light of these new advances including deep reservoir computing as well as combining it with newer gradient-based approaches. We explore this through a dataset applied to the diagnosis of Parkinson’s disease.

Standardizing loT
Abstract

Standardizing loT Security Through the Use of a Dedicated Security Hub

     The ever-increasing popularity of IoT devices has been consistently met with security concerns. By introducing a dedicated device that implements industry-proven security practices between IoT devices and the internet, dependence on manufacturers to properly secure their devices is eliminated. We introduce the concept of the Security Hub, a middleware, to manage all inter-device communication and ensure devices obtain timely updates. Our empirical studies based on a prototype implementation ensured minimal communication overhead and reliable updating of the IoT devices, even under heavy load. With proper manufacturer adoption, this could greatly reduce the security risk of IoT devices on a network.

Predicting Dropout Rate
Abstract

Predicting Students Dropout Rate using Convolutional Neural Networks

    In this paper, we represent student GPA data over four or more semesters using a grayscale image. Each pixel in the image represents a grade point of a student in a given course of a semester. We use Convolutional Neural Networks (CNNs) to analyze that image and determine whether students will finish the program successfully or will drop out prior to program completion. TensorFlow is used to implement the CNN classification technique. The results of this experiment will be useful to instructors and advisors to guide students in course selection. This would help students find their field of interest.

Intelligent Shopping Agent
Abstract

Design and Implementation of Intelligent Shopping Agent using Machine Learning

    With rapid evolution of technology, ever increasing demand for highly personalized shopping. Customers expect their shopping experience to be meaningful. Present e-commerce websites help customers via their search engine recommendations. Our system presents customers with a more tailored and personalized shopping predictions with accuracy when they are ready to shop. In this paper, we discuss the architectural design and implementation of a prototype intelligent shopping assistant that applies machine learning to understand users shopping pattern and predicts their shopping list. The proposed system is implemented as an application using multiple database technologies and Amazon Web Services.

Handwritten Character Rec
Abstract

Applying Context to Handwritten Character Recognition

    Today’s approaches to automate handwritten character recognition utilize various forms of machine learning and while they are, at times, highly accurate, they do not attempt to apply higher-level knowledge to improve performance. This research employs support vector machine (SVM) trained recognizers supplemented with domain knowledge to provide top-down guidance in an attempt to improve recognition accuracy. Results showed SVM recognition alone only provided 60-69% accuracy. With top-down guidance however, accuracy improved to 99.3-99.9%. Modern business uses of smartphones and other camera enabled devices amplify the need for 100% accurate character recognition. This approach brings us closer to that goal.

Improvement of Mobile Applications
Abstract

Energy Efficiency Improvement of Mobile Applications Aided with Edge Computingb>

    Edge computing has been widely adopted as an effective approach to increasing the performance of various websites and applications. Meanwhile. mobile devices are known for their limited computing resource. Computing intensive applications usually have poor performance and even worse, they consume a large portion of battery life. We propose a new mobile application system aided with edge computing so that intensive computing tasks will be executed in the edge. We will demonstrate that not only the application performance will be improved, but the energy efficiency will be significantly increased.

C++ and Python
Abstract

C++ and Python Object Oriented Programs to Perform Data Extraction and Correlation for ISS-CREAM Project

    Cosmic Ray Energetics and Mass (CREAM) is an instrument on the International Space Station designed to detect cosmic rays at high energies. There are multiple input data streams: science data, instrument health monitoring (housekeeping), times of South Atlantic Anomaly passage (SAA), pedestals, and several types of calibration files. Text files are parsed to extract relevant data. We developed programs that use inheritance to correlate the different data streams by timestamp. Our final output is a ROOT tree that incorporates variables from the raw files for each day the instrument sends data.

GIS Spatial
Abstract

GIS Spatial Analyses of Wind and Solar Energy

    Renewable energy compared to non-renewable energy is cleaner for the environment. There are previous studies on renewable energy in the US, however, they were not examining the spatial patterns of the data, only the energy potential in an area. Spatial patterns are valuable to examine. The objective of this study is to complete spatial pattern analyses of the US renewable energy with focus on wind and solar energy. The analysis results show that solar is highly geographically dependent while wind is highly variable. The importance of this study is to have analyses available for possible growth in this industry.

Amazon ECS
Abstract

Amazon Elastic Container Service (ECS)

    This project is about setting up a cluster with fargate technology that can be used with Amazon ECS (Elastic Container Service) to run containers without managing servers or clusters of Amazon EC2 instances. The Amazon ECS creates a task definition that uses fargate launch type, schedules task and configure the cluster in the Amazon ECS. It is also about the detailed study of advantages container has over VM based on the parameters such as density, speed, low management overhead and portability.it also includes the experiments performed to prove the container benefits over VM based on use cases suitable for both.

Edge Computing
Abstract

Edge Computing

    Edge Computing Abstract The goal of the Edge Computing project is to compare the runtimes of the Lambda Management Console, a mobile app, and Edge Computing to try and find that Edge is the software that runs code the fastest and uses the least amount of energy. Edge Computing is a method of allowing applications to be optimized by bringing it closer to the user no matter where the original server is located. For example, if I’m located in Ohio and the original program was uploaded from Moscow, the data would still be brought closer to Ohio where it was created to allow for faster runtimes and less energy consumption. If the comparison shows that Edge is faster than the other options, it increases the viability of using Edge at a higher level.

AWS Lambda
Abstract

Edge Computing

    Edge Computing Abstract The goal of the Edge Computing project is to compare the runtimes of the Lambda Management Console, a mobile app, and Edge Computing to try and find that Edge is the software that runs code the fastest and uses the least amount of energy. Edge Computing is a method of allowing applications to be optimized by bringing it closer to the user no matter where the original server is located. For example, if I’m located in Ohio and the original program was uploaded from Moscow, the data would still be brought closer to Ohio where it was created to allow for faster runtimes and less energy consumption. If the comparison shows that Edge is faster than the other options, it increases the viability of using Edge at a higher level.

Accelerating Serverless Application
Abstract

Serverless Architecture

    Serverless compute service such as AWS Lambda has been widely adopted by software developers today as it eliminates server configuration and server maintenance. AWS Lambda@Edge is an integration of AWS Lambda and AWS CloudFront that allows AWS functions to execute closer to the user and web objects to be cached. This research will demonstrate how AWS Lambda@Edge can be used to improve the performance of a web application when cached objects are delivered. User experience will also be enhanced as web content can be customized based on user’s preference such as languages.

Twitter Streaming