Yash Shah
@ynshah.bsky.social
PhD student at Stanford. Self-proclaimed computational neuroscientist and humanist. Incomplete bio at https://ynshah3.github.io/.
And R-MDN makes equitable predictions across population groups, such as across both boys and girls when performing sex classification on the ABCD (Casey et al., 2008) dataset in the presence of pubertal development scores as the confounder. [9/n]
July 13, 2025 at 9:08 PM
And R-MDN makes equitable predictions across population groups, such as across both boys and girls when performing sex classification on the ABCD (Casey et al., 2008) dataset in the presence of pubertal development scores as the confounder. [9/n]
R-MDN can also remove the influence from multiple confounding variables, as seen when testing on the ADNI (Mueller et al., 2005) dataset. [8/n]
July 13, 2025 at 9:07 PM
R-MDN can also remove the influence from multiple confounding variables, as seen when testing on the ADNI (Mueller et al., 2005) dataset. [8/n]
Since R-MDN is a normalization layer, it can be tacked on to various already-proposed model architectures. [7/n]
July 13, 2025 at 9:07 PM
Since R-MDN is a normalization layer, it can be tacked on to various already-proposed model architectures. [7/n]
R-MDN effectively removes confounder influence from learned DNN features, as rigorously verified in both synthetically controlled environments and real-world datasets. [6/n]
July 13, 2025 at 9:07 PM
R-MDN effectively removes confounder influence from learned DNN features, as rigorously verified in both synthetically controlled environments and real-world datasets. [6/n]
I am excited to share that my work on "Confounder-Free Continual Learning via Recursive Feature Normalization" has been accepted at ICML 2025! Very grateful to @camgonza.bsky.social , @mhabbasi.bsky.social , @qingyuz.bsky.social, Kilian Pohl, and @eadeli.bsky.social for always supporting me. [1/n]
July 13, 2025 at 9:04 PM
I am excited to share that my work on "Confounder-Free Continual Learning via Recursive Feature Normalization" has been accepted at ICML 2025! Very grateful to @camgonza.bsky.social , @mhabbasi.bsky.social , @qingyuz.bsky.social, Kilian Pohl, and @eadeli.bsky.social for always supporting me. [1/n]