Test-Time Training with Masked Autoencoders

Code

• github.com/yossigandelsman/test_time_training_mae
  Officialpytorch★ 94
• github.com/Rima-ag/TTT-MAE
  pytorch★ 0

Abstract

Test-time training adapts to a new test distribution on the fly by optimizing a model for each test input using self-supervision. In this paper, we use masked autoencoders for this one-sample learning problem. Empirically, our simple method improves generalization on many visual benchmarks for distribution shifts. Theoretically, we characterize this improvement in terms of the bias-variance trade-off.

Reproductions