1 paper
Bhushan Chaudhary, Anubha Pandey, Deepak Bhatt +1
Addressing bias in the trained machine learning system often requires access to sensitive attributes. In practice, these attributes are not available either due to legal and policy…