Multi-domain learning by meta-learning: Taking optimal steps in multi-domain loss landscapes by inner-loop learning

Anthony Sicilia, Xingchen Zhao, Davneet S. Minhas, Erin E. O'Connor, Howard J. Aizenstein, William E. Klunk, Dana L. Tudorascu, Seong Jae Hwang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We consider a model-agnostic solution to the problem of Multi-Domain Learning (MDL) for multi-modal applications. Many existing MDL techniques are model-dependent solutions which explicitly require nontrivial architectural changes to construct domain-specific modules. Thus, properly applying these MDL techniques for new problems with well-established models, e.g. U-Net for semantic segmentation, may demand various low-level implementation efforts. In this paper, given emerging multi-modal data (e.g., various structural neuroimaging modalities), we aim to enable MDL purely algorithmically so that widely used neural networks can trivially achieve MDL in a model-independent manner. To this end, we consider a weighted loss function and extend it to an effective procedure by employing techniques from the recently active area of learning-to-learn (meta-learning). Specifically, we take inner-loop gradient steps to dynamically estimate posterior distributions over the hyperparameters of our loss function. Thus, our method is model-agnostic, requiring no additional model parameters and no network architecture changes; instead, only a few efficient algorithmic modifications are needed to improve performance in MDL. We demonstrate our solution to a fitting problem in medical imaging, specifically, in the automatic segmentation of white matter hyperintensity (WMH). We look at two neuroimaging modalities (T1-MR and FLAIR) with complementary information fitting for our problem.

Original languageEnglish
Title of host publication2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021
PublisherIEEE Computer Society
Pages650-654
Number of pages5
ISBN (Electronic)9781665412469
DOIs
Publication statusPublished - 2021 Apr 13
Event18th IEEE International Symposium on Biomedical Imaging, ISBI 2021 - Nice, France
Duration: 2021 Apr 132021 Apr 16

Publication series

NameProceedings - International Symposium on Biomedical Imaging
Volume2021-April
ISSN (Print)1945-7928
ISSN (Electronic)1945-8452

Conference

Conference18th IEEE International Symposium on Biomedical Imaging, ISBI 2021
Country/TerritoryFrance
CityNice
Period21/4/1321/4/16

Bibliographical note

Funding Information:
This work was supported by the NIH/NIA (R01 AG063752, RF1 AG025516, P01 AG025204, K23 MH118070), and SCI Undergraduate Research Scholars Award. We report no conflicts of interests.

Publisher Copyright:
© 2021 IEEE.

All Science Journal Classification (ASJC) codes

  • Biomedical Engineering
  • Radiology Nuclear Medicine and imaging

Fingerprint

Dive into the research topics of 'Multi-domain learning by meta-learning: Taking optimal steps in multi-domain loss landscapes by inner-loop learning'. Together they form a unique fingerprint.

Cite this