We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Fair Generative Models via Transfer Learning

Abstract: This work addresses fair generative models. Dataset biases have been a major cause of unfairness in deep generative models. Previous work had proposed to augment large, biased datasets with small, unbiased reference datasets. Under this setup, a weakly-supervised approach has been proposed, which achieves state-of-the-art quality and fairness in generated samples. In our work, based on this setup, we propose a simple yet effective approach. Specifically, first, we propose fairTL, a transfer learning approach to learn fair generative models. Under fairTL, we pre-train the generative model with the available large, biased datasets and subsequently adapt the model using the small, unbiased reference dataset. We find that our fairTL can learn expressive sample generation during pre-training, thanks to the large (biased) dataset. This knowledge is then transferred to the target model during adaptation, which also learns to capture the underlying fair distribution of the small reference dataset. Second, we propose fairTL++, where we introduce two additional innovations to improve upon fairTL: (i) multiple feedback and (ii) Linear-Probing followed by Fine-Tuning (LP-FT). Taking one step further, we consider an alternative, challenging setup when only a pre-trained (potentially biased) model is available but the dataset that was used to pre-train the model is inaccessible. We demonstrate that our proposed fairTL and fairTL++ remain very effective under this setup. We note that previous work requires access to the large, biased datasets and is incapable of handling this more challenging setup. Extensive experiments show that fairTL and fairTL++ achieve state-of-the-art in both quality and fairness of generated samples. The code and additional resources can be found at bearwithchris.github.io/fairTL/.
Comments: Accepted in AAAI-2023
Subjects: Machine Learning (cs.LG); Computers and Society (cs.CY)
Cite as: arXiv:2212.00926 [cs.LG]
  (or arXiv:2212.00926v1 [cs.LG] for this version)

Submission history

From: Christopher Teo [view email]
[v1] Fri, 2 Dec 2022 01:44:38 GMT (1442kb,D)

Link back to: arXiv, form interface, contact.