SPRING 2018

Melanoma Detection: Computer Vision for Skin Cancer

Kian Ghodoussi,  Leena Mathur,  Priyank Aranke,  Zane Durante

Early detection, diagnosis, and prevention of melanoma skin cancer is crucial for patient survival. Far too often, melanoma progresses to advanced stages because patients lack the ability to self-diagnose malignant moles and do not feel the need to consult a dermatologist until it becomes too late. We are using a style transfer algorithm to create a unique dataset of cell-phone quality, labelled, melanoma mole images. Using this data, we will be training a convolutional neural network to predict the likelihood of moles being cancerous, in order to inspire patients to pursue further treatment and medical consultation to reduce the likelihood of fatal melanoma.

2018 National Data Science Bowl: Cell Nucleus Segmentation

Rachit Kataria,  Lucas Hu,  Sean Syed

Identifying cell nuclei, which hold the cell’s DNA, is the starting point for most disease analysis. Identifying nuclei allows researchers to identify each individual cell in a sample, and by measuring how cells react to various treatments, researchers can understand the underlying biological processes at work. By using deep learning-based image segmentation techniques, we hope to automate the process of identifying nuclei, which will allow for more efficient drug testing, shortening the 10 years it takes for each new drug to come to market.

Kawasaki Disease: Rare Disease Diagnosis using Machine Learning (1.0)

Kawasaki Disease is a rare heart disease that affects children all over the world; however, there is currently no successful diagnostic test for the disease. This means that Kawasaki Disease can often be left undiagnosed, sometimes with fatal consequences. We aim to use machine learning techniques such as SVMs, Boosted Decision Trees, and Deep Neural Networks to create a robust diagnosis algorithm that learns from its mistakes and helps save lives.

LA City Budget: Understanding Civic Spending

Kian Ghodoussi * 
* Project Lead

Every year, the LA budget team manually goes through the year’s budget appropriations and manually categorizes each budget entry; this categorization makes it easier to aggregate the city’s spending by category later on. However, since new entries may not perfectly corresponding to existing, pre-categorized budget entries, it takes not only large amounts of work, but also subjective institutional knowledge to correctly categorize each entry. For this reason, many entries simply go uncategorized year-to-year. We are currently working with the LA Mayor’s Office to apply natural language processing techniques to automatically categorize new budget entries, making it understand LA’s annual spending, and to hopefully enable additional insights on how this spending could be improved.

Sports Sentiment Analytics

Shomik Jain * 
* Project Lead

Sporting events are often an emotional rollercoaster, with fans often voicing opinions and reactions on Social Media. Using Machine Learning and Natural Language Processing techniques such as Sentiment Analysis, our aim is to study sentiment surrounding teams and players, and to see whether any correlation can be found between sentiment, game predictions by fans, and actual game outcomes.

Music Generation

Applying artificial intelligence to create music has been a popular research task for many years. However, recent deep learning approaches have allowed for significant improvement in this domain and have opened many potential avenues for this task. We are working to further explore music creation through the deep learning lense as well as developing methods that use artificial intelligence to assist with the creative process. Current projects include developing algorithms to assist in the creation of jazz heads (sheet music that includes a melody and the corresponding chords) and using Generative Adversarial Networks to develop different styles of drum beats.