In representation learning, there has been recent interest in developing algorithms to disentangle the ground-truth generative factors behind a dataset, and metrics to quantify how fully this occurs. However, these algorithms and metrics often assume that both representations and ground-truth factors are flat, continuous, and factorized, whereas many real-world generative processes involve rich hierarchical structure, mixtures of discrete and continuous variables with dependence between them, and even varying intrinsic dimensionality. In this work, we develop benchmarks, algorithms, and metrics for learning such hierarchical representations.

Ensembles depend on diversity for improved performance. Many ensemble training methods, therefore, attempt to optimize for diversity, which they almost always define in terms of differences in training set predictions. In this paper, however, we demonstrate the diversity of predictions on the training set does not necessarily imply diversity under mild covariate shift, which can harm generalization in practical settings. To address this issue, we introduce a new diversity metric and associated method of training ensembles of models that extrapolate differently on local patches of the data manifold. Across a variety of synthetic and real-world tasks, we find that our method improves generalization and diversity in qualitatively novel ways, especially under data limits and covariate shift.

Deep neural networks have proven remarkably effective at solving many classification problems, but have been criticized recently for two major weaknesses: the reasons behind their predictions are uninterpretable, and the predictions themselves can often be fooled by small adversarial perturbations. These problems pose major obstacles for the adoption of neural networks in domains that require security or transparency. In this work, we evaluate the effectiveness of defenses that differentiably penalize the degree to which small changes in inputs can alter model predictions. Across multiple attacks, architectures, defenses, and datasets, we find that neural networks trained with this input gradient regularization exhibit robustness to transferred adversarial examples generated to fool all of the other models. We also find that adversarial examples generated to fool gradient-regularized models fool all other models equally well, and actually lead to more "legitimate," interpretable misclassifications as rated by people (which we confirm in a human subject experiment). Finally, we demonstrate that regularizing input gradients makes them more naturally interpretable as rationales for model predictions. We conclude by discussing this relationship between interpretability and robustness in deep neural networks.

We often desire our models to be interpretable as well as accurate. Prior work on optimizing models for interpretability has relied on easy-to-quantify proxies for interpretability, such as sparsity or the number of operations required. In this work, we optimize for interpretability by directly including humans in the optimization loop. We develop an algorithm that minimizes the number of user studies to find models that are both predictive and interpretable and demonstrate our approach on several data sets. Our human subjects results show trends towards different proxy notions of interpretability on different datasets, which suggests that different proxies are preferred on different tasks.

Expressive classifiers such as neural networks are among the most accurate supervised learning methods in use today, but their opaque decision boundaries make them difficult to trust in critical applications. We propose a method to explain the predictions of any differentiable model via the gradient of the class label with respect to the input (which provides a normal to the decision boundary). Not only is this approach orders of magnitude faster at identifying input dimensions of high sensitivity than sample-based perturbation methods (e.g. LIME), but it also lends itself to efficiently discovering multiple qualitatively different decision boundaries as well as decision boundaries that are consistent with expert annotation. On multiple datasets, we show our approach generalizes much better when test conditions differ from those in training.