The interaction between people and social media through machine learning and AI algorithms poses a challenge. Despite the widespread coverage of news articles and events influenced by social media, a continuous echo chamber grows due to machine learning algorithms developed by some of the most prominent corporations globally. These algorithms are utilized to exhibit user-generated content through advertisements. Facebook, for instance, had 1.35 billion monthly active users in September 2014 and is predicted to earn 9 billion in global Facebook advertising revenue in 2015 (Duffett, 2015). This predicament has adversely impacted the general public and the billions of individuals who use apps like Facebook, Instagram, and Twitter, creating walls around the content only an individual can see. One probable cause of this issue is the malicious intent of marketing and monetization targeting specific groups of people. A possible solution to this problem could involve conducting a study to investigate how these algorithms operate and why they are precise across several groups of people.

When singing from sheet music, how does the individual know where to start the base note? Indian classical music’s answer to this question is shruti or tonic. A performer selects a tonic (base pitch) based on their vocal range, and that tonic is used throughout their performance. In order to create automated systems which can understand Indian classical music, information about tonic is vital. This research focuses on Mel Frequency Cepstral Coefficients (MFCCs), a popular feature in speech recognition and music information retrieval. Using a correlational research approach on a database of over 1,000 audio samples, this project aims to find correlational indicators of tonic in the MFCCs.

Climate change forces society to increase energy efficiency and stop burning fossil fuels. Buildings cause 36% of global energy use annually, and low-income residential buildings are particularly source due to older, more inefficient technology. Many technologies exist to solve this problem, such as air-to-water heat pumps and thermostat schedules. Furthermore, public and private sector incentives make efficiency upgrades free or very low cost. Despite these incentives, the need for energy retrofits that improve efficiency remains. This research project aims to find the root cause of the lack of progress and determine the best way to correct it. Data collected through interviews with experts in this field will be analyzed to produce the results. A Likert scale will assess the research to determine efficacy methods for improving building energy consumption.

The increased carbon dioxide in the atmosphere due to anthropogenic activities is causing global warming and affecting all life on earth. Climate change is happening faster than species can adapt, and the decline of terrestrial carbon sinks will make the situation worse. Mycorrhizal fungi’s symbiotic relationship with plants is crucial for terrestrial ecosystem functioning and soil organic carbon storage. A meta-analysis found that elevated CO2 can enhance carbon sinks through Arbuscular mycorrhiza (AM) symbiosis, offsetting 9.5 Gt to 15 Gt CO2e by 2100 and reducing the global temperature by 0.5℃. AM-fungi can potentially increase the land carbon sink and solve the global environmental crisis.

Type 1 diabetes affects nearly 1.45 million Americans, most of whom have to use insulin therapy as a treatment, which is inefficient and can lead to diabetic complications. This project investigates the efficacy of gene editing without immunosuppression as a treatment for type 1 diabetes.
Comparative research will review previous examinations and scientific studies of treatments for factors such as blood sugar levels, hyperglycemic episodes, and overall health. The calculating average blood sugar and hyperglycemic episodes for each treatment group will determine if there is a necessary need for expanding research in this field and if gene editing without immunosuppression is an effective treatment for type 1 diabetes.

The potential for excellence in sports remains vastly untapped due to several factors, including the limitations of current recruitment methods. Many talented individuals, who may be overlooked or undervalued due to circumstances beyond their control, are hindered from reaching their full potential. Artificial intelligence (AI) has the potential to revolutionize the sports industry by improving recruitment methods and identifying previously overlooked talent. This project aims to explore the use of AI in sports recruitment through various research methods, including content analysis and evaluation research. The goal is to examine the impact of AI in other industries and apply those findings to the sports industry, with the ultimate goal of elevating the level of play and increasing the potential for success in sports.

Across the globe, human activity's impact on marine ecosystems has been increasingly harmful. Living in the Bay Area means that this impact is more direct, as it is easier for the effects of human civilization to reach marine wildlife. Seabirds are essential indicators of marine ecosystem health; since they typically occupy high trophic levels, such as secondary or tertiary consumers, their varied diet reflects the health of the entire food web. Over the last few decades, the health of marine ecosystems has been declining drastically (by about 69.7% from 1950 to 2010). Due to this decline in health, there have been efforts worldwide to reduce human impact; a large part of this is the implementation of Marine Protected Areas (MPAs). The effectiveness of these areas has long been disputed; this study aims to determine whether MPAs help preserve biodiversity in the Bay Area and, if so, to what extent.

Since the 1970s, the sharp increase in fishing pressure on the shark populations of California has been the subject of much concern. For leopard sharks, one of the most common sharks in San Francisco Bay, protective measures might’ve been too little too late, as the species is vulnerable to outside pressures. In the three years between 1980 and 1983, reported commercial landings of leopard sharks increased from 40,085 lbs to 101,309 lbs, predominantly in San Francisco Bay. In response to the expansion of leopard shark fisheries, the 1992 and 1993 California regulations put in place a daily possession limit of up to three sharks at or over 36 inches; however, many researchers have calculated that this is not enough for a positive increase in population growth. A species as vulnerable to exploitation and extinction as the leopard shark deserves an examination and comparison of multiple different conservation measures to most effectively provide for a stable and expanding population, a task this research will attempt to complete through the use of a Leslie Matrix model to empirically estimate the theoretical population growth if each method was used, allowing for a direct comparison and a solution to the question of how to best protect this species.

Aortic aneurysms (AAs) are localized dilations of the aortic wall, prone to rupture with expansion, an often deadly event (Chaikof et al., 2018). The decision to intervene upon an AA is based primarily upon standardized size criteria and rate of expansion over time as determined radiographically. Early detection and surveillance are essential to timely intervention. This study proposes an end-to-end deep learning algorithm for sequential slice-by-slice segmentation of AAs in computed tomography angiography (CTA). The algorithm uses predicted masks to compute the size of AAs through contour area measurements. The goal of sequential quantification of AA sizes in a series of single-slice CTA images is to identify the largest aorta slice when measuring the maximum diameter. Current manual methods of inspection have led to large amounts of intra-operator variability in measuring maximum aortic diameter among radiologists (Cayne et al., 2004). Image data for training the algorithm was acquired in an unconventional web-sourced manner from open case studies and preprints.

Despite the increase in accessibility of efforts to improve individuals’ well-being, mental health has become an increasingly problematic issue in the United States. Notably, following the COVID-19 pandemic, a large proportion of high school students have reported frequently experiencing feelings of sadness and hopelessness, in line with the deterioration of mental health, every few weeks. The methods proposed to solve this glaring issue rely on helpful, though highly rudimentary, telemedicine services and face-to-face sessions with psychologists. While effective, these sessions tend to be limited by their in-person nature and obstructed by financial barriers, compounding student-held stigma. This research project investigates the effectiveness of novel statistics and machine learning advances in quickly discerning anxiety and depression, two of the most pertinent mental health illnesses among young adults. Through the correlative analysis of the Substance Abuse and Mental Health Services Administration’s (SAMHSA) mental health (MH-CLD) datasets from 2019 and 2021, the project aims to create a machine learning framework, taking into account demographic background and “life stressors,” to recognize these illnesses with the Keras API.