5153 research outputs found
Sort by
Maimonides Risk Parity
No full textPhilip Maymin's poster examining Maimonides risk parity and regular risk parity
Study of deep learning approaches to face recognition and object detection YOLO
No full textIn this research, I have focused on deep learning approaches to face detection and recognition and object detection and recognition. This research has mainly focused on training the neural networks or other models with enough amounts of data so that it achieves desirable results. Starting with the basics of neural network where a neuron, the smallest unit in deep learning field, is defined and explained, I have elevated the research to the topic where I could recognize the face of a person or an object in an image. First the neural networks have been introduced in this research and then training of the neural networks has been attained with both the CPU and the GPU. In an algorithm called matrix form of back propagation, the use of multiple GPUs has been made along with CUDA kernel and cuBLAS library. Application of face recognition has been implemented using the pre trained model Facenet and Deep Convolutional Neural Networks. After analyzing neural networks and its facial recognition application, other approaches of deep learning have been given force to. I have used the library OpenCV along with deep learning approaches to implement face recognition, Image registration and YOLO Object Detection and Recognition with the tensor flow and keras environment supported by anaconda for python
Prep Adherence to Tenofovir-Based Therapeutics Seems Efficacious in Reducing HIV Infection Dissemination: A Comparative Evaluation on The Therapeutic Potency of Tenofovir-Based Prep & Dapivirine-Based Antiretroviral (ARV) Microbicide in HIV Prevention
No full textThis dissertation is being archived for campus access to current students and staff only at the request of the author.HIV/AIDS is a serious public health issue causing millions of morbidities and thousands of mortalities with the immense socio-economic burden. This disease adversely impacts population science on a global scale. The development of an effective HIV vaccine remains a therapeutic challenge. Two principles pre-exposure prophylaxis (PrEP) approaches using Tenofovir-based, and Dapivirine-based therapeutic regimens are compared to HIV preventive efficacy in high-risk populations. This dissertation aims to determine if PrEP adherence to Tenofovir-based therapeutics as single or combined treatment approaches are more efficacious than Dapivirine microbicides-based therapeutic approaches in HIV prevention in high-risk populations. The trial outcomes are methodically segmented in four different therapeutic clusters of i) TDF-FTC, ii) TDF-alone, iii) 1% TFV gel, and iv) DVR to determine HIV preventive efficacy from individual treatment cluster, targeting 26,755 HIV uninfected high-risk populations. The cumulative HIV preventive therapeutic potency is assessed on a PICO table and measured conducting box plot and waterfall analyses. The results of this analysis demonstrated a 55% risk reduction from TDF-FTC combined treatment, 34% from TDF single treatment, 27% from TFV single treatment and 42% risk reduction from DVR Microbicides ring single treatment clusters. Men achieved the highest HIV preventive efficacy at 86% from the TDF-FTC group, whereas women achieved 71% of HIV preventive efficacy from the TDF-alone group
Music Industry and Blockchain based smart contract
No full textThis poster looks at the impact of blockchain on musical industry and focuses on the impact of blockchain on technological artists. In looking through the sector's supply chain, we argue that on-demand streaming (such as Spotify and Apple Music) platforms have made it easy for consumers to access music but have still created an interaction between artists and customers that results in the inefficiency of royalty payment. Its purpose is to identify blockchain applications that enable artists to disintegrate the industry and create more value from their own products. This paper examines some applications and concepts related to blockchain, including smart agreements, record keeping, revenue management and analytics of metadata. By introducing examples, we evaluate how companies market this new model and how these models are confined to the current state of the art in the music industry
Using Signal Processing Techniques to Detect Sleep Apnea Events
No full textNowadays, with the development of modern technology, we can detect sleep apnea by using Electroencephalography (EEG), Electrocardiography (ECG), blood pressure (BP), and Respiration rates (RR). These signals are recorded in time domain, however for us to extract vital information we must analyze them in frequency domain. This research will examine sleep apnea; how to understand it and in doing so determine the best methods of detection using several different signal processing techniques. By doing so we hope to help physicians further improve their detection methods as well as their accuracy
An Idealized Scenario for Energy Generation by Nuclear Fusion
No full textWe study nuclear fusion processes in a deuteron plasma under a combination of conditions such that, for a given energy input, a maximum energy output can be attained. Specifically we consider fusion processes initiated by the rapid adiabatic compression by a piston of a deuteron plasma contained in a well‐insulated chamber. To exploit the n2 factor in the fusion reaction rate, we consider one mole of plasma which, at ambient temperature and pressure, provides a particle density of ~ 10^19 cm^‐3. Reaction rates are enhanced by the application of magnetic and electric fields to reduce the degrees of freedom of the plasma, thereby lowering its heat capacity and producing a higher temperature increase for a given energy input. Computations show that the combination of adiabatic operation, high particle density and reduced degrees of freedom can result in appreciable fusion rates at temperatures lower than those in magnetic confinement experiments. We consider both primary D-D and secondary D-T reactions. Conditions of energy break-even were found at temperatures of the order of 10^6 K
Intelligent Multi-Attribute Decision Making Applications: Decision Support Systems for Performance Measurement, Evaluation and Benchmarking
No full textEfficiency has been and continues to be an important attribute of competitive business environments where limited resources exist. Owing to growing complexity of organizations and more broadly, to global economic growth, efficiency considerations are expected to remain a top priority for organizations. Continuous performance evaluations play a significant role in sustaining efficient and effective business processes. Consequently, the literature offers a wide range of performance evaluation methodologies to assess the operational efficiency of various industries. Majority of these models focus solely on quantitative criteria omitting qualitative data. However, a thorough performance measurement and benchmarking require consideration of all available information since accurately describing and defining complex systems require utilization of both data types. Most evaluation models also function under the unrealistic assumption of evaluation criteria being dependent on one another. Furthermore, majority of these methodologies tend to utilize discrete and contemporary information eliminating historical performance data from the model environment. These shortcomings hinder the reliability of evaluation outcomes leading to inadequate performance evaluations for many businesses. This problem gains more significance for business where performance evaluations are tied in to important decisions relating to business expansion, investment, promotion and compensation. The primary purpose of this research is to present a thorough, equitable and accurate evaluation framework for operations management while filling the existing gaps in the literature. Service industry offers a more suitable platform for this study since the industry tend to accommodate both qualitative and quantitative performance evaluation factors relatively with more ease compared to manufacturing due to the intensity of customer (consumer) interaction. Accordingly, a U.S. based food franchise company is utilized for data acquisition and as a case study to demonstrate the applications of the proposed models. Compatible with their multiple criteria nature, performance measurement, evaluation and benchmarking systems require heavy utilization of Multi-Attribute Decision Making (MADM) approaches which constitute the core of this research. In order to be able to accommodate the vagueness in decision making, fuzzy values are also utilized in all proposed models. In the first phase of the study, the main and sub-criteria in the evaluation are considered independently in a hierarchical order and contemporary data is utilized in a holistic approach combining three different multi-criteria decision making methods. The cross-efficiency approach is also introduced in this phase. Building on this approach, the second phase considered the influence of the main and sub-criteria over one another. That is, in the proposed models, the main and sub-criteria form a network with dependencies rather than having a hierarchical relationship. The decision making model is built to extract the influential weights for the evaluation criteria. Furthermore, Group Decision Making (GDM) is introduced to integrate different perspectives and preferences of multiple decision makers who are responsible for different functions in the organization with varying levels of impact on decisions. Finally, an artificial intelligence method is applied to utilize the historical data and to obtain the final performance ranking. Owing to large volumes of data emanating from digital sources, current literature offers a variety of artificial intelligence and machine learning methods for big data analytics applications. Comparing the results generated by the ANNs, three additional well-established methods, viz., Adaptive Neuro Fuzzy Inference System (ANFIS), Least Squares Support Vector Machine (LSSVM) and Extreme Learning Machine (ELM), are also employed for the same problem. In order to test the prediction capability of these methods, the most influencing criteria are obtained from the data set via Pearson Correlation Analysis and grey relational analysis. Subsequently, the corresponding parameters in each method are optimized via Particle Swarm Optimization to improve the prediction accuracy. The accuracy of artificial intelligence and machine learning methods are heavily reliant on large volumes of data. Despite the fact that several businesses, especially business that utilize social media data or on-line real-time operational data, there are organizations which lack adequate amount of data required for their performance evaluations simply due to the nature of their business. Grey Modeling (GM) technique addresses this issue and provides higher forecasting accuracy in presence of uncertain and limited data. With this motivation, a traditional multi-variate grey model is applied to predict the performance scores. Improved grey models are also applied to compare the results. Finally, the integration of the fractional order accumulation along with the background value coefficient optimization are proposed to improve accuracy
Smart Automation System
No full textThis Poster involves the study of fault detection and correction in automation industry with the use of PLC, SCADA, GSM and smart sensor technology. Detection of fault can be made easily with the use of SCADA and proper implementation of PLC in automation machines , but this project shows how to control that detected fault with the use of mobile or GSM device and how to get a solution of that fault using smart sensors. This can same time and resources in protecting human lives and machines. The design of this smart system includes smart technology and smart devices. The use of PLC- SCADA -GSM is must but for further improvement, smart technology used
The Influence of President's and Government's Political Manifestos on the Economy
No full textThe data sets of political manifestos will present us with invaluable knowledge of trying to explain some points of micro and macro-economic occurrences in the US. With the researched data set, we will establish whether American leaders and their team had influences on economy. We will observe whether their foresight went as planned of it was just a political game and correlations seemed completely random and chaotic. And maybe we can connect the dots with the politics and the economy, if any correlation is established
Nasa Connecticut Space Grant Consortium: Balloon/Drone-based Aerial Platforms for Remote Particulate Matter Pollutant Monitoring
No full textIn this research a balloon/drone based aerial platform for remote particulate matter pollutant monitoring is developed. Using balloon and drones, it can monitor air pollution at high altitude above ground and trace the origin as needed. Based on Arduino platform, the collected data are sent back to ground station via long-range wireless communication. It is then be uploaded to cloud server so that users can monitor the real-time air pollution remotely from smartphones and laptops. Compared to traditional ground air pollution monitoring stations, such aerial platform can detect air pollution above ground with improved convenience and flexibility. It can also be extended as a general aerial monitoring platform for other environmental and geographical conditions