changed if man file

_posts/2024-10-07-ifman.md

@@ -112,7 +112,7 @@ We also anticipate our attack to work better with smaller training sets, as ther
 
 Recently influence functions have been proposed to increase the fairness of downstream models <d-cite key="li2022achieving, wang2024fairif, sattigeri2022fair, kong2021resolving, pang2024fair, chhabra2023data, chen2024fast, yao2023understanding, ghosh2023biased"></d-cite>, we focus on the study by <d-cite key="li2022achieving"></d-cite> because it uses the same definition of influence as us. In this study, influence scores from a base model are used to increase the fairness of a downstream model. Since fairness of the downstream model is guided by influence scores, an adversary with an incentive to reduce fairness would be interested in manipulating them.
 
-We propose an untargeted attack for this use-case : scale the base model $\theta^{*}$ by a constant $\lambda > 0$. The malicious base model output by the model trainer is now $\theta^\prime = \lambda \cdot \theta^*$, instead of $\theta^*$. Note that for logistic regression the malicious and original base model are indistinguishable since scaling with a positive constant maintains the sign of the predictions, leading to the same accuracy.
+We propose an untargeted attack for this use-case : scale the base model $\theta^*$ by a constant $\lambda > 0$. The malicious base model output by the model trainer is now $\theta^\prime = \lambda \cdot \theta^*$, instead of $\theta^*$. Note that for logistic regression the malicious and original base model are indistinguishable since scaling with a positive constant maintains the sign of the predictions, leading to the same accuracy.
 
 #### Experimental Results