Parameter Tunning and Model Improvement on Efficient Data Deletion in Machine Learning

Yue Lyu, Zhenghua Chen, Xiaobin Shang

Abstract: We validated the work stated in the paper ”Making AI Forget You: Data Deletion in Machine Learning”[1] by reproducing their results. We found that our results roughly aligned the trends of their results but with some minor fluctuations. We further explored their proposed models by fine tuning their hyperparameters. The effects of number of iterations and tree depth on the performance of corresponding proposed models were investigated. We further optimized DC-K-means model and proposed a new competitive weighted DC-K-means model, which has better statistical performance on some datasets at a minor cost of runtime efficiency.

Track: Ablation

NeurIPS Paper Id: https://openreview.net/forum?id=H1xovVBg8B&noteId=SklrKRP7sH