Debugging Machine Learning
Explaining misclassified instances in machine learning
during training to support informed debugging of data and features
Final project for the class Special Topics in Information Visualization, taught by Michelle Borkin at Northeastern University.
Machine Learning algorithms study pattern repetitions to learn and make predictions on data. However, many times the predictions don’t work and the end-users must train the algorithm to correct the labeling. If the output is not interpretable by the users, the task will frustrate them. The project tries to solve this problem through an interactive visualization system that explains the raw data (that were either misclassified or properly labeled).
The website was coded in HTML5, CSS and D3.js
Video presentation of Debugging Machine Learning, done for the class by Farnaz Irannejad
