A data-visualization dashboard combating implicit bias in machine learning models. By visualizing the change in output as a function of training iterations, one can detect (visually and statistically) unexpected data-shifts called drift. Drift is caused by the infinitely many variables that implicitly affect real-world data. Detecting drift combats said bias by exposing flaws that may not have been considered in model development. Additionally, our application detects feature-drift, thereby pinpointing the source(s) of bias.
Contributors:
- Eyasu Woldu
- Lincoln Mcloud
- Chase Gormley
- Kevin Bacon
- Likhon Gomes
