FAIR-TAT: Improving Model Fairness Using Targeted Adversarial Training | IEEE Conference Publication | IEEE Xplore