Hello! I'm excited to share that I've concluded my 10-week internship at Vanderbilt University's Department of Biomedical Informatics. Looking back, it seems like the summer passed so quickly—yet I have learned so much from so many people. I want to thank Kim Unertl & Rischelle Jenkins for organizing this program and Brad Malin, Nicholas Jackson, & Victor Borza for their valuable mentorship this summer! Also, a special shoutout to my fellow interns for making the summer heat of Nashville not only bearable but enjoyable! If you're interested, here's a very, very, very, quick summary of my project this summer. THE PROJECT: Fairness in Dermatology Image Classification In the field of dermatology, physicians perform better on certain skin tones compared to others. Machine learning models also perform better on certain skin tones than others, which may be caused by imbalanced training data. We hypothesized that balancing our data with oversampling techniques and synthetic GAN-generated images would improve performance disparities across skin tones. I built a code framework to train/fine-tune CNN models and evaluate their performance/fairness on test sets (check the GitHub repo in my profile). Turns out that augmenting our training data with synthetic images may increase model performance while maintaining or even improving fairness. Very cool stuff! Reach out with questions!
This is awesome!!
Congratulations! It was nice meeting you!!!
This is so interesting! Congrats!
Amazing! So proud of you!
Congratulations!
So exciting!
Awesome Ethan!!
Congrats, Ethan!
Communication Design and Psychological & Brain Sciences at Washington University In St. Louis
2moGood job Ethan Feng