By default neural networks are not robust to changes in data distribution. This has been demonstrated with simple image corruptions, such as blurring or adding noise, degrading image classification performance. Many methods have been proposed to mitigate these issues but for the most part models are evaluated on single corruptions. In reality, visual space is compositional in nature, that is, that as well as robustness to elemental corruptions, robustness to compositions of corruptions is also needed. In this work we develop a compositional image classification task where, given a few elemental corruptions, models are asked to generalize to compositions of these corruptions. That is, to achieve compositional robustness. We experimentally compare empirical risk minimization with an invariance building pairwise contrastive loss and, counter to common intuitions in domain generalization, achieve only marginal improvements in compositional robustness by encouraging invariance. To move beyond invariance, following previously proposed inductive biases that model architectures should reflect data structure, we introduce a modular architecture whose structure replicates the compositional nature of the task. We then show that this modular approach consistently achieves better compositional robustness than non-modular approaches. We additionally find empirical evidence that the degree of invariance between representations of 'in-distribution' elemental corruptions fails to correlate with robustness to 'out-of-distribution' compositions of corruptions.

Deep Neural Networks (DNNs) often fail in out-of-distribution scenarios. In this paper, we introduce a tool to visualize and understand such failures. We draw inspiration from concepts from neural electrophysiology, which are based on inspecting the internal functioning of a neural networks by analyzing the feature tuning and invariances of individual units. Deep Electrophysiology, in short Deephys, provides insights of the DNN's failures in out-of-distribution scenarios by comparative visualization of the neural activity in in-distribution and out-of-distribution datasets. Deephys provides seamless analyses of individual neurons, individual images, and a set of set of images from a category, and it is capable of revealing failures due to the presence of spurious features and novel features. We substantiate the validity of the qualitative visualizations of Deephys thorough quantitative analyses using convolutional and transformers architectures, in several datasets and distribution shifts (namely, colored MNIST, CIFAR-10 and ImageNet).

Deep neural networks are the default choice of learning models for computer vision tasks. Extensive work has been carried out in recent years on explaining deep models for vision tasks such as classification. However, recent work has shown that it is possible for these models to produce substantially different attribution maps even when two very similar images are given to the network, raising serious questions about trustworthiness. To address this issue, we propose a robust attribution training strategy to improve attributional robustness of deep neural networks. Our method carefully analyzes the requirements for attributional robustness and introduces two new regularizers that preserve a model's attribution map during attacks. Our method surpasses state-of-the-art attributional robustness methods by a margin of approximately 3% to 9% in terms of attribution robustness measures on several datasets including MNIST, FMNIST, Flower and GTSRB.

We propose a new attribution method for neural networks developed using first principles of causality (to the best of our knowledge, the first such). The neural network architecture is viewed as a Structural Causal Model, and a methodology to compute the causal effect of each feature on the output is presented. With reasonable assumptions on the causal structure of the input data, we propose algorithms to efficiently compute the causal effects, as well as scale the approach to data with large dimensionality. We also show how this method can be used for recurrent neural networks. We report experimental results on both simulated and real datasets showcasing the promise and usefulness of the proposed algorithm.