Addressing the Loss-Metric Mismatch with Adaptive Loss Alignment
This addresses a fundamental problem in machine learning training for practitioners, offering a versatile solution that is transferable across tasks.
The paper tackles the mismatch between fixed loss functions and evaluation metrics by meta-learning an adaptive loss function to directly optimize the metric, showing considerable improvements over state-of-the-art methods in metric learning and classification tasks.
In most machine learning training paradigms a fixed, often handcrafted, loss function is assumed to be a good proxy for an underlying evaluation metric. In this work we assess this assumption by meta-learning an adaptive loss function to directly optimize the evaluation metric. We propose a sample efficient reinforcement learning approach for adapting the loss dynamically during training. We empirically show how this formulation improves performance by simultaneously optimizing the evaluation metric and smoothing the loss landscape. We verify our method in metric learning and classification scenarios, showing considerable improvements over the state-of-the-art on a diverse set of tasks. Importantly, our method is applicable to a wide range of loss functions and evaluation metrics. Furthermore, the learned policies are transferable across tasks and data, demonstrating the versatility of the method.