Bias-Mitigation: code and data for "Mitigating Algorithmic Bias in Multiclass CNN Classifications Using Polynomial Causal Modeling"

Code and data supporting the article "Mitigating Algorithmic Bias in Multiclass CNN Classifications Using Polynomial Causal Modeling" (submitted to Discover Computing). Contents: the Python notebook for data preparation, DeepFace emotion labelling, CNN training and prediction; the R notebook for causal bias detection and mitigation; the original prediction file multiclass_data.csv (5,477 test images) on which all results are based; and DeepFace emotion labels for the FairFace images. FairFace images are not redistributed; see the README for the source.