Environment Diversification with Multi-head Neural Network for Invariant Learning
This work addresses the issue of distributional shift robustness in neural networks, which is a critical problem for machine learning practitioners, but it appears incremental as it builds on existing invariant learning research.
The paper tackles the problem of neural network performance degradation due to distribution shifts between training and testing data by proposing EDNIL, an invariant learning framework with a multi-head neural network to absorb data biases, resulting in empirically more robust models against such shifts.
Neural networks are often trained with empirical risk minimization; however, it has been shown that a shift between training and testing distributions can cause unpredictable performance degradation. On this issue, a research direction, invariant learning, has been proposed to extract invariant features insensitive to the distributional changes. This work proposes EDNIL, an invariant learning framework containing a multi-head neural network to absorb data biases. We show that this framework does not require prior knowledge about environments or strong assumptions about the pre-trained model. We also reveal that the proposed algorithm has theoretical connections to recent studies discussing properties of variant and invariant features. Finally, we demonstrate that models trained with EDNIL are empirically more robust against distributional shifts.