Neural networks (NNs) can learn to rely on spurious signals in the training data, leading to poor generalisation. Recent methods tackle this problem by training NNs with additional ground-truth annotations of such signals. These methods may, however, let spurious signals re-emerge in deep convolutional NNs (CNNs). We propose Targeted Activation Penalty (TAP), a new method tackling the same problem by penalising activations to control the re-emergence of spurious signals in deep CNNs, while also lowering training times and memory usage. In addition, ground-truth annotations can be expensive to obtain. We show that TAP still works well with annotations generated by pre-trained models as effective substitutes of ground-truth annotations. We demonstrate the power of TAP against two state-of-the-art baselines on the MNIST benchmark and on two clinical image datasets, using four different CNN architectures.

Multi-context systems (MCS) are a powerful framework for interlinking heterogeneous knowledge sources. They model the flow of information among different reasoning components (called contexts) in a declarative way, using so-called bridge rules, where contexts and bridge rules may be nonmonotonic. We considerably generalize MCS to managed MCS (mMCS): while the original bridge rules can only add information to contexts, our generalization allows arbitrary operations on context knowledge bases to be freely defined, e.g., deletion or revision operators. The paper motivates and introduces the generalized framework and presents several interesting instances. Furthermore, we consider inconsistency management in mMCS and complexity issues.