Filtering by
- All Subjects: Machine Learning
- Creators: Barrett, The Honors College
- Member of: Barrett, The Honors College Thesis/Creative Project Collection
- Resource Type: Text
Classification in machine learning is quite crucial to solve many problems that the world is presented with today. Therefore, it is key to understand one’s problem and develop an efficient model to achieve a solution. One technique to achieve greater model selection and thus further ease in problem solving is estimation of the Bayes Error Rate. This paper provides the development and analysis of two methods used to estimate the Bayes Error Rate on a given set of data to evaluate performance. The first method takes a “global” approach, looking at the data as a whole, and the second is more “local”—partitioning the data at the outset and then building up to a Bayes Error Estimation of the whole. It is found that one of the methods provides an accurate estimation of the true Bayes Error Rate when the dataset is at high dimension, while the other method provides accurate estimation at large sample size. This second conclusion, in particular, can have significant ramifications on “big data” problems, as one would be able to clarify the distribution with an accurate estimation of the Bayes Error Rate by using this method.
This project aims to incorporate the aspect of sentiment analysis into traditional stock analysis to enhance stock rating predictions by applying a reliance on the opinion of various stocks from the Internet. Headlines from eight major news publications and conversations from Yahoo! Finance’s “Conversations” feature were parsed through the Valence Aware Dictionary for Sentiment Reasoning (VADER) natural language processing package to determine numerical polarities which represented positivity or negativity for a given stock ticker. These generated polarities were paired with stock metrics typically observed by stock analysts as the feature set for a Logistic Regression machine learning model. The model was trained on roughly 1500 major stocks to determine a binary classification between a “Buy” or “Not Buy” rating for each stock, and the results of the model were inserted into the back-end of the Agora Web UI which emulates search engine behavior specifically for stocks found in NYSE and NASDAQ. The model reported an accuracy of 82.5% and for most major stocks, the model’s prediction correlated with stock analysts’ ratings. Given the volatility of the stock market and the propensity for hive-mind behavior in online forums, the performance of the Logistic Regression model would benefit from incorporating historical stock data and more sources of opinion to balance any subjectivity in the model.
This project considers the FPGA implementations of MLP and CNN feedforward. While FPGAs provide significant performance improvements, they come at a substantial financial cost. We explore the options of implementing these algorithms on a smaller budget. We successfully implement a multilayer perceptron that identifies handwritten digits from the MNIST dataset on a student-level DE10-Lite FPGA with a test accuracy of 91.99%. We also apply our trained network to external image data loaded through a webcam and a Raspberry Pi, but we observe lower test accuracy in these images. Later, we consider the requirements necessary to implement a more elaborate convolutional neural network on the same FPGA. The study deems the CNN implementation feasible in the criteria of memory requirements and basic architecture. We suggest the CNN implementation on the same FPGA to be worthy of further exploration.
Methods: The standard NLP process was used for this study in which a gold standard was reached through matched paired annotations of the forum text in brat and a neural network was trained on the content. Following the annotation process, adjudication occurred to increase the inter-annotator agreement. Categories were developed by local physicians to describe the questions and three pilots were run to test the best way to categorize the questions.
Results: The inter-annotator agreement, calculated via F-score, before adjudication for a 0.7 threshold was 0.378 for the annotation activity. After adjudication at a threshold of 0.7, the inter-annotator agreement increased to 0.560. Pilots 1, 2, and 3 of the categorization activity had an inter-annotator agreement of 0.375, 0.5, and 0.966 respectively.
Discussion: The inter-annotator agreement of the annotation activity may have been low initially since the annotators were students who may have not been as invested in the project as necessary to accurately annotate the text. Also, as everyone interprets the text slightly differently, it is possible that that contributed to the differences in the matched pairs’ annotations. The F-score variation for the categorization activity partially had to do with different delivery systems of the instructions and partially with the area of study of the participants. The first pilot did not mandate the use of the original context located in brat and the instructions were provided in the form of a downloadable document. The participants were computer science graduate students. The second pilot also had the instructions delivered via a document, but it was strongly suggested that the context be used to gain an understanding of the questions’ meanings. The participants were also computer science graduate students who upon a discussion of their results after the pilot expressed that they did not have a good understanding of the medical jargon in the posts. The final pilot used a combination of students with and without medical background, required to use the context, and included verbal instructions in combination with the written ones. The combination of these factors increased the F-score significantly. For a full-scale experiment, students with a medical background should be used to categorize the questions.
This paper’s field of study falls into the cross section of geology and fire science, history, social conflict, public service ethics, and collaborative failures. I explore how a series of small choices snowballed into a full, government funded relocation effort after attempts at controlling the anthracite coal seam fire failed. Geology and fire science worked in tandem during the mine fire, influencing each other and complicating the firefighting efforts. The fire itself was a unique challenge. The history of Centralia played a large role in the government and community response efforts. I use the borough and regional history to contextualize the social conflict that divided Centralia. Social conflict impaired the community’s ability to unify and form a therapeutic community, and in turn, it damaged community-government relationships. The government agencies involved in the mine fire response did their own damage to community relationships by pursuing their own interests. Agencies worried about their brand image, and politicians worried about re-election. I study how these ethical failures impacted the situation. Finally, I look at a few examples of collaborative failures on behalf of the government and the community. Over the course of my research, it became apparent the people killed Centralia, not the fire.
Meanwhile, machine learning is growing more advanced by the day. Online retailers like Amazon run complex algorithms to recommend future purchases and monitor price changes. Mobile phones use neural networks to interpret speech. GPS apps track anonymous motion data in smartphones to give real-time traffic estimates. Artificial intelligence is becoming increasingly ubiquitous because of its versatility in analyzing and solving human problems; it follows, then, that a machine could learn how to teach humans skills and techniques. HelperBot is a platform fighting game project that employs this cutting-edge learning technology to close the skill gap between novice and veteran gamers as quickly and seamlessly as possible.
This paper analyzes responses to deviated Trolley Problem scenarios [5] in a simulated driving environment and still images from MIT’s moral machine website [8] to better understand how humans respond to various crashes. Also included is participants driving habits and personal values, however the bulk of that analysis is not included here. The results of the simulation prove that for the most part in driving scenarios, people would rather sacrifice themselves over people outside of the vehicle. The moral machine scenarios prove that self-sacrifice changes as the trend to harm one’s own vehicle was not so strong when passengers were introduced. Further defending this idea is the importance placed on Family Security over any other value.
Suggestions for implementing ethics into autonomous vehicle crashes stem from the results of this experiment but are dependent on more research and greater sample sizes. Once enough data is collected and analyzed, a moral baseline for human’s moral domain may be agreed upon, quantified, and turned into hard rules governing how self-driving cars should act in different scenarios. With these hard rules as boundary conditions, artificial intelligence should provide training and incremental learning for scenarios which cannot be determined by the rules. Finally, the neural networks which make decisions in artificial intelligence must move from their current “black box” state to something more traceable. This will allow researchers to understand why an autonomous vehicle made a certain decision and allow tweaks as needed.