• Gender beyond the binary: computationally mapping gender to a spectrum using sex differences in the brain

      Williams, Reed (2022-05)
      Biological sex is far more complex than simply two categories: male and female. The mere existence of transgender and intersex individuals displays this complexity clearly on the surface, while the differences between cisgender people within their own respective categories brings this idea to a deeper level. While sex differences reveal themselves in many different scientific disciplines, this study will focus on findings in the field of neuroscience; specifically, it will narrow in on volumetric measurements of brain regions known to have differing trends across the male and female sexes. The construction of a surrogate data set driven by measurements extracted from existing literature will be used to fit a logistic regression model. The resulting probability function will be used to first create a base Biological Sex Spectrum; this refers to a representation of biological sex as a spectrum in the absence of societal influence. This probability function will then be modified to produce a Societally Influenced Gender Spectrum; this refers to a spectrum that has been influenced by the concept of the gender binary and more closely represents our current world. The comparison of these two spectra will reveal the space for an increase in gender diversity as societal views continue shifting further away from restricting gender stereotypes.
    • A generative chatbot with natural language processing

      Liebman, David (2020-12)
      The goal in this thesis is to create a chatbot, a computer program that can respond verbally to a human in the course of simple day-to-day conversations. A deep learning neural network model called the Transformer is used to develop the chatbot. A full description of a Transformer is provided. The use of a few different Transformer-based Natural Language Processing models to develop the chatbot, including Generative Pre-Training 2 (GPT2), are shown. For comparison a Gated Recurrent Unit (GRU) based model is included. Each of these are explained below. The chatbot code is installed on a small device such as the Raspberry Pi with speech recognition and speech-to-text software. In this way a device that can carry out a verbal conversation with a human might be created. For the GRU-based model a Raspberry Pi 3B with 1GB RAM can be used. A Raspberry Pi 4B with 4GB of RAM is needed to run a chatbot with the GPT2.
    • A healthcare IoT prototype for responsive oxygen therapy treatment of COPD patients

      Khan, Azer (2020-12)
      Our final design offers oxygen therapy patients an IoT enabled, adaptable and small form-factor device offering potential for automatic detection and agile response to oxygen saturation readings. Our key components are a photodiode sensor, algorithm processor and micro- controller providing the foundation for future development for FDA approval, machine-learning and analytics, and feedback-loop oxygen tank controller tracks. Our device is cutting-edge in its communications, power consumption and efficiency. We have gained an understanding of the effort required to design an IoT enabled solution in the healthcare space. Integrating hardware and software designs are an exercise to understand the inner workings of many systems used today. It is estimated that there will be about 50 billion IoT-enabled devices in the world by 2030 [22]. Technologists who understand the underlying systems will be able to make well-informed decisions about the future of the connected world.
    • Raspberry pi embedded operating system and runtime

      Perry, James J. (2016-05)
      This thesis explores the creation of a small footprint, high-performance Embedded Operating System (EOS) for the Raspberry Pi (RPi). Using a customization approach, the image is configures to include only required functions and omits nonessential functions. The result preserves available memory and storage for use during runtime of an embedded solution. As part of this process, the thesis leverages the resulting runtime environment to provide complex functions (i.e. inter process messaging and GPIO support) that run atomically (noninterruptible).