Search Constraints
Filtering by:
Department
Computer Science
Remove constraint Department: Computer Science
Collection
Thesis
Remove constraint Collection: Thesis
« Previous |
1 - 10 of 993
|
Next »
Number of results to display per page
Search Results
- Creator:
- Zheng, Huiping
- Description:
- The Efficient Market Hypothesis (EMH) says that there is no better forecast of stock price possible. If one believes in the EMH, then the stock price movements should follow a random walk. Stock price changes should be random and unpredictable. Despite this hypothesis, we decided to build three back-propagation neural networks and three recurrent neural networks to forecast the daily closing price of stock indexes (S&P500, Dow Jones, and NASDAQ). Our experiments showed that the price is predictable and much better than the random guess. Using a simple shortterm investment strategy, a good annual profit rate can be obtained. Different activation functions and different data preprocessing techniques were tested in order to dynamically determine the best neural network topology. We explored more than six hundred network structures for each neural network in our experiments. The same data sets were also analyzed by statistical models using a commercial statistics package. We find that the parameters of the statistical models can help us in determining the recurrent neural network structure. Comparing the forecasting results, our recurrent neural networks outperform the statistical models in terms of the annual profit rates.
- Resource Type:
- Thesis
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Jaramillo, Richard
- Description:
- The optimization of lossy compression of images is inherently difficult due to the complexity of human visual perception. Existing optimization techniques perform adequately, but often stop at local optimal points in the search space of compression solutions. Evolutionary programming is an alternative to human-designed algorithms that has been applied successfully to overcome local optima in machine learning applications. Evolutionary programming borrows from nature the concepts of survival-of-the-fittest and the crossover of genetic material between individuals in a population to produce a new generation of individuals. The result is a guided search through the solution space for a problem. Genetic algorithms (GA) and genetic programming (GP) are two types of evolutionary programming. GA has been applied to many applications including vector quantization (VQ) compression codebook generation. GP is an extension of genetic algorithms in which the individuals in the evolving population are represented by hierarchically structured program trees instead of fixed length strings of characters. GP has been applied in various fields such as analog circuit design with results that rival human designed solutions to the same problems. No previous research has been found that applies GP to the problem of VQ codebook generation. Co-evolution is an extension of GP in which a population of problems to be solved evolves at the same time as the solution population evolves. Co-evolution has been shown to perform better than normal evolution at overcoming local optima for some applications. This research examines a novel use of genetic programming and coevolutionary genetic programming for the evolution and co-evolution of VQ codebooks for image compression. The results show that codebooks generated using traditional human-coded compression algorithms can be further optimized with the use of GP and co-evolutionary GP.
- Resource Type:
- Thesis
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Bhosale, Radhika
- Description:
- In our previous study we have proposed ResNet-based network models, named ContiNet and F-ContiNet, for sign language recognition. ContiNet is a 3D convolutional network with a constant number of temporal snapshots, and F-ContiNet is an evolution of ContiNet where the fusion idea is applied so that the spatial and temporal information from the early layers will have greater direct impact on the final recognition. We evaluated the models with RGB (raw) and Mask (Skeleton) video inputs from a Chinese sign language (CSL) dataset. The experiment results indicate that the proposed ContiNet and F-ContiNet outperformed the state-of-the-art approaches, and taking Mask videos as inputs always resulted in better performance. The purpose of this project as reported here is to further our understanding of the ContiNet and F-ContiNet models using a different dataset. We used the same experiment settings to evaluate ContiNet and F-ContiNet over the Microsoft sign language (MSL) dataset. To our surprise, the experiment results were opposite of the findings on the CSL dataset. In particular, ContiNet and F-ContiNet did not produce the best performance, and taking RGB videos as inputs resulted in better performance in most settings. We have analyzed the quality of the MSL dataset to diagnose the problems and possible solutions.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Xu, Runyu
- Description:
- Machine Learning (ML) becomes increasingly popular; industry spends billions of dollars building ML systems. Data scientists have come up with many good algorithms and trained models. However, putting those ML models into production is still in the early stage. The deployment process is distinct from that for traditional software applications; it is not yet well understood among data scientists and IT engineers in their roles and responsibilities, resulting in many anti-pattern practices [21]. The key issues identified by researchers at Google[40] include lack of production-like prototyping stack for data scientists, monolithic programs not fitted for component based ML system orchestration, and lack of best practices in system design. To find solutions, teams need to understand the inherent structure of ML systems and to find ML engineering best practices. This paper presents an abstraction of ML system design process, a design pattern named Model-Service-Client + Retraining (MSC/R) consisting of four main components: Model (data and trained model), Service (model serving infrastructure), Client (user interface), and Retraining (model monitoring and retraining). Data scientists and engineers can use this pattern as a discipline in designing and deploying ML pipelines methodically. They can separate concerns, modularize ML systems, and work in parallel. This paper also gives case studies on how to use MSC/R to quickly and reliably deploy two ML models -- YOLOv3, an object detection model, and Stock Prediction using Long Short-Term Memory (LSTM) algorithm onto AWS and GCP clouds. Two different implementation approaches are used: serving the model as a microservice RESTful API on AWS managed container platform ECS, and on GCP serverless platform Cloud Run. In the end, this paper gives analysis and discussion on how using the MSC/R design pattern helps to meet the objectives of implementing ML production systems and solve the common problems. It also provides insights and recommendations.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Supekar, Vishakha
- Description:
- The changing environment plays a vital role in the health of humans as well as animals. Due to the grave impacts of pollution and increased temperature on human health, it is of utmost importance to monitor the environment parameters at every step. With the recent developments in the Internet of Things (IoT), monitoring these parameters in real-time has become possible and cost-effective. In this project, I have designed and implemented a web application to demonstrate an intelligent temperature and humidity reporting system using IoT and cloud. This application aims to provide a visual map for the users to analyze the temperatures in different areas to make an informed decision. This system can be used as a prototype in strengthening real-time temperature data and humidity data in many applications such as Nest, activating farm sprinklers based on weather data, and helping patients sensitive to high temperature and high humidity to take prompt action upon real-time notification on change in temperature or humidity. The proposed system provides a new solution by utilizing the sensor activity on various applications as it is represented using Amazon Web Services IoT, which is an emerging area of research.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Peng, GueiSian
- Description:
- We analyze the efficiency of the state-of-the-art, object detection systems that have recently been introduced for fast and accurate object detection in images, video streams and real-time videos. We implemented and analyzed the efficiency of YOLOv2, YOLOv3 and SSD object detection systems. In this project, we introduce the functionality of these 3 systems, the metrics to evaluate the efficiency of object detection algorithms and present the results of our implementations for small-scale datasets. We also present an efficiency analysis of these three systems for large-scale datasets. Object detection is an intelligent computer vision technique, similar to our humans’ visions, for locating instances of objects in images, video or real-time surveillance. It has been researched for several years and has been improved to an unprecedented level. It also has been adopted across our daily lives from our cellphones, video surveillance, and object tracking to pedestrian recognition and so forth. There are various detection models such as Region Proposals (R-CNN, Fast R-CNN, Faster R-CNN), You Only Look Once (YOLO), Single Shot MultiBox Detector (SSD), etc. We studied the most recent and most efficient systems: YOLOv2, YOLOv3, and SSD. YOLOv3 and SSD are considered the best methods with the highest accuracy and fastest speed that can be achieved in object detection in images and video streams. We also studied the efficiency of all three systems for real-time object detections. The results demonstrated that YOLOv3 is the most accurate but slowest object detection system while SSD is the fastest one with the lowest accuracy. YOLOv2 has a lower accuracy than YOLOv3 but it is faster. For object detection in recorded images and videos, YOLOv3 is the best one since it detects the objects with the highest accuracy, however for real-time video-streams, SSD provides the best one since it is the fastest one. Since there is a trade-off between accuracy and speed in all these systems, the most appropriate system for each application depends on the application requirements.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Sharma, Nishtha
- Description:
- Large organizations like a university typically have many web applications deployed and used by a larger number of users every day. As such an application changes, it should be thoroughly tested to avoid unexpected service failures or interruptions. The purpose of this project is to evaluate the pros and cons of the Page-Object design pattern in Selenium-based testing of web applications. We have experimented with two web applications deployed at CSUSM and our conclusion is two-fold. On the one hand the Page-Object design pattern can largely improve readability due to clean separation between page-specific test code and production code. On the other hand, it improves the change management of large web applications: any UI change can easily be implemented, updated and maintained into the Page Objects.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Zheng, Huiping
- Description:
- The Efficient Market Hypothesis (EMH) says that there is no better forecast of stock price possible. If one believes in the EMH, then the stock price movements should follow a random walk. Stock price changes should be random and unpredictable. Despite this hypothesis, we decided to build three back-propagation neural networks and three recurrent neural networks to forecast the daily closing price of stock indexes (S&P500, Dow Jones, and NASDAQ). Our experiments showed that the price is predictable and much better than the random guess. Using a simple shortterm investment strategy, a good annual profit rate can be obtained. Different activation functions and different data preprocessing techniques were tested in order to dynamically determine the best neural network topology. We explored more than six hundred network structures for each neural network in our experiments. The same data sets were also analyzed by statistical models using a commercial statistics package. We find that the parameters of the statistical models can help us in determining the recurrent neural network structure. Comparing the forecasting results, our recurrent neural networks outperform the statistical models in terms of the annual profit rates.
- Resource Type:
- Thesis
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Tomar, Anjali
- Description:
- The use of version control has gained a huge spread over the past decade. This has led to their use in almost every industry. Whether it is a software company, private organization or educational institution all are using version control for a variety of reasons. Specifically it is very useful in the software development industry for developers, who makes various changes almost every day as a part of the software development process. That's why I choose version control as a research topic for my project. I am proposing to build an application to enhance the better use of a type of version control, i.e., GitHub .The application I have created is a website. The web application I developed is a nice tool to fetch GitHub repository metadata of GitHub’s public repositories. Application needs to give a project & repository name to get metadata. Application also helps to analyses user commits in the repository using Dashboard. The main purpose of the application is easy access to everyone, because it does not require email or password from the user. Any user can access the website and its functionality. This project will aid the easy to use environment by improved engagement among students and professors, and also software professionals and non-professionals.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science
- Creator:
- Maya, Manuel
- Description:
- Technology is a tool used to help make societies daily lives easier. The personal computer helped pave the path for technological advancements. First came the personal computer, then smart phones came as the next major advancement and now there is the emergence of Internet of Things. Internet of Things are smaller embedded computers used to connect to a computer or smart phone. Internet of Things are used to collect, monitor, display and analyze information. Currently, of the most popular Internet of Things devices are: Amazon’s Alexa Echo Dot, Nest Smart Thermometer, Google Home, Wyze Security Cameras, Smart Baby Monitors and Phillips Hue Lightbulbs. These devices all have the ability to monitor and display information through either a web portal, mobile or computer application. Internet of Things devices have been ridiculed for being in people’s homes and gathering information. Although there have been security concerns about hackers breaking into these Internet of Things devices and stealing sensitive information, they are not the only security threat; big companies are also as well. Companies have been using user information for targeted ads. A large number of users have not known their information is being used for this reason. This is why it is vital for users to know what information their Internet of Things devices are capturing without their consent. This can be done by capturing network traffic via a Raspberry Pi running Kali Linux using Aircrack-ng, and then analyzing the packets using Wireshark. User information should be private and protected and no one should have their information taken from them without their consent. That is why it is also vital to have some type of monitoring device to see if a malicious attacker is trying to steal your data. Not only should this monitoring program create a database of the previous attacks that happened in the user’s network, but also inform the user in real time.
- Resource Type:
- Project
- Campus Tesim:
- San Marcos
- Department:
- Computer Science