Selective Classification for Deep Neural Networks
This enables DNNs to be used in mission-critical applications by providing guaranteed error rates, though it is an incremental adaptation of existing selective classification techniques to deep learning.
The paper tackles the problem of improving deep neural network prediction reliability by introducing selective classification, which rejects uncertain instances to meet a user-specified risk level, achieving a guaranteed 2% error in top-5 ImageNet classification with 99.9% probability and 60% coverage.
Selective classification techniques (also known as reject option) have not yet been considered in the context of deep neural networks (DNNs). These techniques can potentially significantly improve DNNs prediction performance by trading-off coverage. In this paper we propose a method to construct a selective classifier given a trained neural network. Our method allows a user to set a desired risk level. At test time, the classifier rejects instances as needed, to grant the desired risk (with high probability). Empirical results over CIFAR and ImageNet convincingly demonstrate the viability of our method, which opens up possibilities to operate DNNs in mission-critical applications. For example, using our method an unprecedented 2% error in top-5 ImageNet classification can be guaranteed with probability 99.9%, and almost 60% test coverage.