We algorithmically identify label errors in the test sets of 10 of the most commonly-used computer vision, natural language, and audio datasets, and subsequently study the potential for these label errors to affect benchmark results. Errors in test sets are numerous and widespread: we estimate an average of 3.4% errors across the 10 datasets, where for example 2916 label errors comprise 6% of the ImageNet validation set. Putative label errors are found using confident learning and then human-validated via crowdsourcing (54% of the algorithmically-flagged candidates are indeed erroneously labeled). Surprisingly, we find that lower capacity models may be practically more useful than higher capacity models in real-world datasets with high proportions of erroneously labeled data. For example, on ImageNet with corrected labels: ResNet-18 outperforms ResNet-50 if the prevalence of originally mislabeled test examples increases by just 6%. On CIFAR-10 with corrected labels: VGG-11 outperforms VGG-19 if the prevalence of originally mislabeled test examples increases by 5%. Traditionally, ML practitioners choose which model to deploy based on test accuracy -- our findings advise caution here, proposing that judging models over correctly labeled test sets may be more useful, especially for noisy real-world datasets.
Conditional quantile estimation is a key statistical learning challenge motivated by the need to quantify uncertainty in predictions or to model a diverse population without being overly reductive. As such, many models have been developed for this problem. Adopting a meta viewpoint, we propose a general framework (inspired by neural network optimization) for aggregating any number of conditional quantile models in order to boost predictive accuracy. We consider weighted ensembling strategies of increasing flexibility where the weights may vary over individual models, quantile levels, and feature values. An appeal of our approach is its portability: we ensure that estimated quantiles at adjacent levels do not cross by applying simple transformations through which gradients can be backpropagated, and this allows us to leverage the modern deep learning toolkit for building quantile ensembles. Our experiments confirm that ensembling can lead to big gains in accuracy, even when the constituent models are themselves powerful and flexible.
Reliant on too many experiments to learn good actions, current Reinforcement Learning (RL) algorithms have limited applicability in real-world settings, which can be too expensive to allow exploration. We propose an algorithm for batch RL, where effective policies are learned using only a fixed offline dataset instead of online interactions with the environment. The limited data in batch RL produces inherent uncertainty in value estimates of states/actions that were insufficiently represented in the training data. This leads to particularly severe extrapolation when our candidate policies diverge from one that generated the data. We propose to mitigate this issue via two straightforward penalties: a policy-constraint to reduce this divergence and a value-constraint that discourages overly optimistic estimates. Over a comprehensive set of 32 continuous-action batch RL benchmarks, our approach compares favorably to state-of-the-art methods, regardless of how the offline data were collected.
Automated machine learning (AutoML) can produce complex model ensembles by stacking, bagging, and boosting many individual models like trees, deep networks, and nearest neighbor estimators. While highly accurate, the resulting predictors are large, slow, and opaque as compared to their constituents. To improve the deployment of AutoML on tabular data, we propose FAST-DAD to distill arbitrarily complex ensemble predictors into individual models like boosted trees, random forests, and deep networks. At the heart of our approach is a data augmentation strategy based on Gibbs sampling from a self-attention pseudolikelihood estimator. Across 30 datasets spanning regression and binary/multiclass classification tasks, FAST-DAD distillation produces significantly better individual models than one obtains through standard training on the original data. Our individual distilled models are over 10x faster and more accurate than ensemble predictors produced by AutoML tools like H2O/AutoSklearn.
While image classification models have recently continued to advance, most downstream applications such as object detection and semantic segmentation still employ ResNet variants as the backbone network due to their simple and modular structure. We present a simple and modular Split-Attention block that enables attention across feature-map groups. By stacking these Split-Attention blocks ResNet-style, we obtain a new ResNet variant which we call ResNeSt. Our network preserves the overall ResNet structure to be used in downstream tasks straightforwardly without introducing additional computational costs. ResNeSt models outperform other networks with similar model complexities. For example, ResNeSt-50 achieves 81.13% top-1 accuracy on ImageNet using a single crop-size of 224x224, outperforming previous best ResNet variant by more than 1% accuracy. This improvement also helps downstream tasks including object detection, instance segmentation and semantic segmentation. For example, by simply replace the ResNet-50 backbone with ResNeSt-50, we improve the mAP of Faster-RCNN on MS-COCO from 39.3% to 42.3% and the mIoU for DeeplabV3 on ADE20K from 42.1% to 45.1%.
We present TraDE, an attention-based architecture for auto-regressive density estimation. In addition to a Maximum Likelihood loss we employ a Maximum Mean Discrepancy (MMD) two-sample loss to ensure that samples from the estimate resemble the training data. The use of attention means that the model need not retain conditional sufficient statistics during the process beyond what is needed for each covariate. TraDE performs significantly better than existing approaches such differentiable flow based estimators on standard tabular and image-based benchmarks in terms of the log-likelihood on held out data. TraDE works well wide range of tasks that includes classification methods to ascertain the quality of generated samples, out of distribution sample detection, and handling outliers in the training data.
Image classifiers are typically scored on their test set accuracy, but high accuracy can mask a subtle type of model failure. We find that high scoring convolutional neural networks (CNN) exhibit troubling pathologies that allow them to display high accuracy even in the absence of semantically salient features. When a model provides a high-confidence decision without salient supporting input features we say that the classifier has overinterpreted its input, finding too much class-evidence in patterns that appear nonsensical to humans. Here, we demonstrate that state of the art neural networks for CIFAR-10 and ImageNet suffer from overinterpretation, and find CIFAR-10 trained models make confident predictions even when 95% of an input image has been masked and humans are unable to discern salient features in the remaining pixel subset. Although these patterns portend potential model fragility in real-world deployment, they are in fact valid statistical patterns of the image classification benchmark that alone suffice to attain high test accuracy. We find that ensembling strategies can help mitigate model overinterpretation, and classifiers which rely on more semantically meaningful features can improve accuracy over both the test set and out-of-distribution images from a different source than the training data.