Mitigating Real-World Distribution Shifts in the Fourier Domain

Page view(s)
4
Checked on Apr 06, 2024
Mitigating Real-World Distribution Shifts in the Fourier Domain
Title:
Mitigating Real-World Distribution Shifts in the Fourier Domain
Journal Title:
Transactions on Machine Learning Research
DOI:
Publication Date:
02 August 2023
Citation:
Krishnamachari, K., Ng, S., & Foo, C. (2023). Mitigating Real-World Distribution Shifts in the Fourier Domain. Trans. Mach. Learn. Res., 2023.
Abstract:
While machine learning systems can be highly accurate in their training environments, their performance in real-world deployments can suffer significantly due to distribution shifts. Real-world distribution shifts involve various input distortions due to noise, weather, device and other variations. Many real-world distribution shifts are not represented in standard domain adaptation datasets and prior empirical work has shown that domain adaptation methods developed using these standard datasets may not generalize well to real-world distribution shifts. Furthermore, motivated by observations of the sensitivity of deep neural networks (DNN) to the spectral statistics of data, which can vary in real-world scenarios, we propose Fourier Moment Matching (FMM), a model-agnostic input transformation that matches the Fourier-amplitude statistics of source to target data using unlabeled samples. We demonstrate through extensive empirical evaluations across time-series, image classification and semantic segmentation tasks that FMM is effective both individually and when combined with a variety of existing methods to overcome real-world distribution shifts.
License type:
Publisher Copyright
Funding Info:
There was no specific funding for the research done
Description:
ISSN:
2835-8856
Files uploaded:

File Size Format Action
fmm-tmlr-1.pdf 3.06 MB PDF Open