Towards Trustworthy Predictions from Deep Neural Networks with Fast Adversarial Calibration
This work is significant for AI systems guiding decision-making in real-world applications, as it improves the trustworthiness of predictions under domain shift, which is a common challenge in deployment.
The paper addresses the problem of obtaining trustworthy and well-calibrated predictions from deep neural networks, especially under domain shift. It introduces a training strategy combining an entropy-encouraging loss with an adversarial calibration loss, which substantially outperforms existing state-of-the-art approaches in yielding well-calibrated predictions under various domain drifts.
To facilitate a wide-spread acceptance of AI systems guiding decision making in real-world applications, trustworthiness of deployed models is key. That is, it is crucial for predictive models to be uncertainty-aware and yield well-calibrated (and thus trustworthy) predictions for both in-domain samples as well as under domain shift. Recent efforts to account for predictive uncertainty include post-processing steps for trained neural networks, Bayesian neural networks as well as alternative non-Bayesian approaches such as ensemble approaches and evidential deep learning. Here, we propose an efficient yet general modelling approach for obtaining well-calibrated, trustworthy probabilities for samples obtained after a domain shift. We introduce a new training strategy combining an entropy-encouraging loss term with an adversarial calibration loss term and demonstrate that this results in well-calibrated and technically trustworthy predictions for a wide range of domain drifts. We comprehensively evaluate previously proposed approaches on different data modalities, a large range of data sets including sequence data, network architectures and perturbation strategies. We observe that our modelling approach substantially outperforms existing state-of-the-art approaches, yielding well-calibrated predictions under domain drift.