arXiv (Cornell University) · 2021 · 25 citations · 0 references
ComputingComputer ScienceComputer VisionMachine LearningOptimizationDeep LearningGradient-based OptimizationLoss FunctionOptimization For Machine LearningRepresentation LearningSupervised LearningData DistributionDomain GeneralizationEvaluation ProtocolsFisher InformationRegularizationRegularization TechniquesRobust LossRobust Loss Functions
Learning robust models that generalize well under changes in the data distribution is critical for real-world applications. To this end, there has been a growing surge of interest to learn simultaneously from multiple training domains - while enforcing different types of invariance across those domains. Yet, all existing approaches fail to show systematic benefits under controlled evaluation protocols. In this paper, we introduce a new regularization - named Fishr - that enforces domain invariance in the space of the gradients of the loss: specifically, the domain-level variances of gradients are matched across training domains. Our approach is based on the close relations between the gradient covariance, the Fisher Information and the Hessian of the loss: in particular, we show that Fishr eventually aligns the domain-level loss landscapes locally around the final weights. Extensive experiments demonstrate the effectiveness of Fishr for out-of-distribution generalization. Notably, Fishr improves the state of the art on the DomainBed benchmark and performs consistently better than Empirical Risk Minimization. Our code is available at https://github.com/alexrame/fishr.