Abstract
Learning meaningful representations that disentangle the underlying structure
of the data generating process is considered to be of key importance in machine
learning. While disentangled representations were found to be useful for
diverse tasks such as abstract reasoning and fair classification, their
scalability and real-world impact remain questionable. We introduce a new
high-resolution dataset with 1M simulated images and over 1,800 annotated
real-world images of the same setup. In contrast to previous work, this new
dataset exhibits correlations, a complex underlying structure, and allows to
evaluate transfer to unseen simulated and real-world settings where the encoder
i) remains in distribution or ii) is out of distribution. We propose new
architectures in order to scale disentangled representation learning to
realistic high-resolution settings and conduct a large-scale empirical study of
disentangled representations on this dataset. We observe that disentanglement
is a good predictor for out-of-distribution (OOD) task performance.
Original language | English |
---|---|
Title of host publication | Proceedings of International Conference on Learning Representations |
Number of pages | 22 |
Publication date | 2021 |
Publication status | Published - 2021 |
Event | 9th International Conference on Learning Representations - Virtual event Duration: 3 May 2021 → 7 May 2021 |
Conference
Conference | 9th International Conference on Learning Representations |
---|---|
Location | Virtual event |
Period | 03/05/2021 → 07/05/2021 |