TLXML: Task-Level Explanation of Meta-Learning via Influence Functions
This work addresses the need for explainability in meta-learning, which is crucial for real-world applications to mitigate risks of inappropriate model updates, but it is incremental as it adapts existing influence function methods to the meta-learning context.
The paper tackles the problem of explaining meta-learning models by proposing influence functions to measure how training tasks affect adaptation and inference, demonstrating its effectiveness through experiments on task distinction and distribution distinction with MAML and Prototypical Networks on image classification tasks.
The scheme of adaptation via meta-learning is seen as an ingredient for solving the problem of data shortage or distribution shift in real-world applications, but it also brings the new risk of inappropriate updates of the model in the user environment, which increases the demand for explainability. Among the various types of XAI methods, establishing a method of explanation based on past experience in meta-learning requires special consideration due to its bi-level structure of training, which has been left unexplored. In this work, we propose influence functions for explaining meta-learning that measure the sensitivities of training tasks to adaptation and inference. We also argue that the approximation of the Hessian using the Gauss-Newton matrix resolves computational barriers peculiar to meta-learning. We demonstrate the adequacy of the method through experiments on task distinction and task distribution distinction using image classification tasks with MAML and Prototypical Network.