To date, there has been no formal study of the statistical cost of interpretability in machine learning. As such, the discourse around potential trade-offs is often informal and misconceptions abound. In this work, we aim to initiate a formal study of these trade-offs. A seemingly insurmountable roadblock is the lack of any agreed upon definition of interpretability. Instead, we propose a shift in perspective. Rather than attempt to define interpretability, we propose to model the \emph{act} of \emph{enforcing} interpretability. As a starting point, we focus on the setting of empirical risk minimization for binary classification, and view interpretability as a constraint placed on learning. That is, we assume we are given a subset of hypothesis that are deemed to be interpretable, possibly depending on the data distribution and other aspects of the context. We then model the act of enforcing interpretability as that of performing empirical risk minimization over the set of interpretable hypotheses. This model allows us to reason about the statistical implications of enforcing interpretability, using known results in statistical learning theory. Focusing on accuracy, we perform a case analysis, explaining why one may or may not observe a trade-off between accuracy and interpretability when the restriction to interpretable classifiers does or does not come at the cost of some excess statistical risk. We close with some worked examples and some open problems, which we hope will spur further theoretical development around the tradeoffs involved in interpretability.
One of the principal scientific challenges in deep learning is explaining generalization, i.e., why the particular way the community now trains networks to achieve small training error also leads to small error on held-out data from the same population. It is widely appreciated that some worst-case theories -- such as those based on the VC dimension of the class of predictors induced by modern neural network architectures -- are unable to explain empirical performance. A large volume of work aims to close this gap, primarily by developing bounds on generalization error, optimization error, and excess risk. When evaluated empirically, however, most of these bounds are numerically vacuous. Focusing on generalization bounds, this work addresses the question of how to evaluate such bounds empirically. Jiang et al. (2020) recently described a large-scale empirical study aimed at uncovering potential causal relationships between bounds/measures and generalization. Building on their study, we highlight where their proposed methods can obscure failures and successes of generalization measures in explaining generalization. We argue that generalization measures should instead be evaluated within the framework of distributional robustness.
Recent work has explored the possibility of pruning neural networks at initialization. We assess proposals for doing so: SNIP (Lee et al., 2019), GraSP (Wang et al., 2020), SynFlow (Tanaka et al., 2020), and magnitude pruning. Although these methods surpass the trivial baseline of random pruning, they remain below the accuracy of magnitude pruning after training, and we endeavor to understand why. We show that, unlike pruning after training, accuracy is the same or higher when randomly shuffling which weights these methods prune within each layer or sampling new initial values. As such, the per-weight pruning decisions made by these methods can be replaced by a per-layer choice of the fraction of weights to prune. This property undermines the claimed justifications for these methods and suggests broader challenges with the underlying pruning heuristics, the desire to prune at initialization, or both.
The dominant term in PAC-Bayes bounds is often the Kullback--Leibler divergence between the posterior and prior. For so-called linear PAC-Bayes risk bounds based on the empirical risk of a fixed posterior kernel, it is possible to minimize the expected value of the bound by choosing the prior to be the expected posterior, which we call the oracle prior on the account that it is distribution dependent. In this work, we show that the bound based on the oracle prior can be suboptimal: In some cases, a stronger bound is obtained by using a data-dependent oracle prior, i.e., a conditional expectation of the posterior, given a subset of the training data that is then excluded from the empirical risk term. While using data to learn a prior is a known heuristic, its essential role in optimal bounds is new. In fact, we show that using data can mean the difference between vacuous and nonvacuous bounds. We apply this new principle in the setting of nonconvex learning, simulating data-dependent oracle priors on MNIST and Fashion MNIST with and without held-out data, and demonstrating new nonvacuous bounds in both cases.
The information-theoretic framework of Russo and J. Zou (2016) and Xu and Raginsky (2017) provides bounds on the generalization error of a learning algorithm in terms of the mutual information between the algorithm's output and the training sample. In this work, we study the proposal, by Steinke and Zakynthinou (2020), to reason about the generalization error of a learning algorithm by introducing a super sample that contains the training sample as a random subset and computing mutual information conditional on the super sample. We first show that these new bounds based on the conditional mutual information are tighter than those based on the unconditional mutual information. We then introduce yet tighter bounds, building on the "individual sample" idea of Bu, S. Zou, and Veeravalli (2019) and the "data dependent" ideas of Negrea et al. (2019), using disintegrated mutual information. Finally, we apply these bounds to the study of Langevin dynamics algorithm, showing that conditioning on the super sample allows us to exploit information in the optimization trajectory to obtain tighter bounds based on hypothesis tests.
In this work, we focus on the use of influence functions to identify relevant training examples that one might hope "explain" the predictions of a machine learning model. One shortcoming of influence functions is that the training examples deemed most "influential" are often outliers or mislabelled, making them poor choices for explanation. In order to address this shortcoming, we separate the role of global versus local influence. We introduce RelatIF, a new class of criteria for choosing relevant training examples by way of an optimization objective that places a constraint on global influence. RelatIF considers the local influence that an explanatory example has on a prediction relative to its global effects on the model. In empirical evaluations, we find that the examples returned by RelatIF are more intuitive when compared to those found using influence functions.
We introduce "instability analysis," a framework for assessing whether the outcome of optimizing a neural network is robust to SGD noise. It entails training two copies of a network on different random data orders. If error does not increase along the linear path between the trained parameters, we say the network is "stable." Instability analysis reveals new properties of neural networks. For example, standard vision models are initially unstable but become stable early in training; from then on, the outcome of optimization is determined up to linear interpolation. We leverage instability analysis to examine iterative magnitude pruning (IMP), the procedure underlying the lottery ticket hypothesis. On small vision tasks, IMP finds sparse "matching subnetworks" that can train in isolation from initialization to full accuracy, but it fails to do so in more challenging settings. We find that IMP subnetworks are matching only when they are stable. In cases where IMP subnetworks are unstable at initialization, they become stable and matching early in training. We augment IMP to rewind subnetworks to their weights early in training, producing sparse subnetworks of large-scale networks, including Resnet-50 for ImageNet, that train to full accuracy. This submission subsumes 1903.01611 ("Stabilizing the Lottery Ticket Hypothesis" and "The Lottery Ticket Hypothesis at Scale").
We propose to study the generalization error of a learned predictor $\hat h$ in terms of that of a surrogate (potentially randomized) classifier that is coupled to $\hat h$ and designed to trade empirical risk for control of generalization error. In the case where $\hat h$ interpolates the data, it is interesting to consider theoretical surrogate classifiers that are partially derandomized or rerandomized, e.g., fit to the training data but with modified label noise. We show that replacing $\hat h$ by its conditional distribution with respect to an arbitrary $\sigma$-field is a viable method to derandomize. We give an example, inspired by the work of Nagarajan and Kolter (2019), where the learned classifier $\hat h$ interpolates the training data with high probability, has small risk, and, yet, does not belong to a nonrandom class with a tight uniform bound on two-sided generalization error. At the same time, we bound the risk of $\hat h$ in terms of a surrogate that is constructed by conditioning and shown to belong to a nonrandom class with uniformly small generalization error.
In this work, we improve upon the stepwise analysis of noisy iterative learning algorithms initiated by Pensia, Jog, and Loh (2018) and recently extended by Bu, Zou, and Veeravalli (2019). Our main contributions are significantly improved mutual information bounds for Stochastic Gradient Langevin Dynamics via data-dependent estimates. Our approach is based on the variational characterization of mutual information and the use of data-dependent priors that forecast the mini-batch gradient based on a subset of the training samples. Our approach is broadly applicable within the information-theoretic framework of Russo and Zou (2015) and Xu and Raginsky (2017). Our bound can be tied to a measure of flatness of the empirical risk surface. As compared with other bounds that depend on the squared norms of gradients, empirical investigations show that the terms in our bounds are orders of magnitude smaller.