Data deletion algorithms aim to remove the influence of deleted data points from trained models at a cheaper computational cost than fully retraining those models. However, for sequences of deletions, most prior work in the non-convex setting gives valid guarantees only for sequences that are chosen independently of the models that are published. If people choose to delete their data as a function of the published models (because they don't like what the models reveal about them, for example), then the update sequence is adaptive. In this paper, we give a general reduction from deletion guarantees against adaptive sequences to deletion guarantees against non-adaptive sequences, using differential privacy and its connection to max information. Combined with ideas from prior work which give guarantees for non-adaptive deletion sequences, this leads to extremely flexible algorithms able to handle arbitrary model classes and training methodologies, giving strong provable deletion guarantees for adaptive deletion sequences. We show in theory how prior work for non-convex models fails against adaptive deletion sequences, and use this intuition to design a practical attack against the SISA algorithm of Bourtoule et al. [2021] on CIFAR-10, MNIST, Fashion-MNIST.

This work explores the hypothesis that the complexity of the function a deep neural network (NN) is learning can be deduced by how fast its weights change during training. Our analysis provides evidence for this supposition by relating the network's distribution of Lipschitz constants (i.e., the norm of the gradient at different regions of the input space) during different training intervals with the behavior of the stochastic training procedure. We first observe that the average Lipschitz constant close to the training data affects various aspects of the parameter trajectory, with more complex networks having a longer trajectory, bigger variance, and often veering further from their initialization. We then show that NNs whose biases are trained more steadily have bounded complexity even in regions of the input space that are far from any training point. Finally, we find that steady training with Dropout implies a training- and data-dependent generalization bound that grows poly-logarithmically with the number of parameters. Overall, our results support the hypothesis that good training behavior can be a useful bias towards good generalization.

In the recent advances of natural language processing, the scale of the state-of-the-art models and datasets is usually extensive, which challenges the application of sample-based explanation methods in many aspects, such as explanation interpretability, efficiency, and faithfulness. In this work, for the first time, we can improve the interpretability of explanations by allowing arbitrary text sequences as the explanation unit. On top of this, we implement a hessian-free method with a model faithfulness guarantee. Finally, to compare our method with the others, we propose a semantic-based evaluation metric that can better align with humans' judgment of explanations than the widely adopted diagnostic or re-training measures. The empirical results on multiple real data sets demonstrate the proposed method's superior performance to popular explanation techniques such as Influence Function or TracIn on semantic evaluation.

We extend semi-supervised learning to the problem of domain adaptation to learn significantly higher-accuracy models that train on one data distribution and test on a different one. With the goal of generality, we introduce AdaMatch, a method that unifies the tasks of unsupervised domain adaptation (UDA), semi-supervised learning (SSL), and semi-supervised domain adaptation (SSDA). In an extensive experimental study, we compare its behavior with respective state-of-the-art techniques from SSL, SSDA, and UDA on vision classification tasks. We find AdaMatch either matches or significantly exceeds the state-of-the-art in each case using the same hyper-parameters regardless of the dataset or task. For example, AdaMatch nearly doubles the accuracy compared to that of the prior state-of-the-art on the UDA task for DomainNet and even exceeds the accuracy of the prior state-of-the-art obtained with pre-training by 6.4% when AdaMatch is trained completely from scratch. Furthermore, by providing AdaMatch with just one labeled example per class from the target domain (i.e., the SSDA setting), we increase the target accuracy by an additional 6.1%, and with 5 labeled examples, by 13.6%.

SpeechBrain is an open-source and all-in-one speech toolkit. It is designed to facilitate the research and development of neural speech processing technologies by being simple, flexible, user-friendly, and well-documented. This paper describes the core architecture designed to support several tasks of common interest, allowing users to naturally conceive, compare and share novel speech processing pipelines. SpeechBrain achieves competitive or state-of-the-art performance in a wide range of speech benchmarks. It also provides training recipes, pretrained models, and inference scripts for popular speech datasets, as well as tutorials which allow anyone with basic Python proficiency to familiarize themselves with speech technologies.

Pretrained Transformer-based models were reported to be robust in intent classification. In this work, we first point out the importance of in-domain out-of-scope detection in few-shot intent recognition tasks and then illustrate the vulnerability of pretrained Transformer-based models against samples that are in-domain but out-of-scope (ID-OOS). We empirically show that pretrained models do not perform well on both ID-OOS examples and general out-of-scope examples, especially on fine-grained few-shot intent detection tasks. To figure out how the models mistakenly classify ID-OOS intents as in-scope intents, we further conduct analysis on confidence scores and the overlapping keywords and provide several prospective directions for future work. We release the relevant resources to facilitate future research.

Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well on few-shot learning, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class.

Object-based attention is a key component of the visual system, relevant for perception, learning, and memory. Neurons tuned to features of attended objects tend to be more active than those associated with non-attended objects. There is a rich set of models of this phenomenon in computational neuroscience. However, there is currently a divide between models that successfully match physiological data but can only deal with extremely simple problems and models of attention used in computer vision. For example, attention in the brain is known to depend on top-down processing, whereas self-attention in deep learning does not. Here, we propose an artificial neural network model of object-based attention that captures the way in which attention is both top-down and recurrent. Our attention model works well both on simple test stimuli, such as those using images of handwritten digits, and on more complex stimuli, such as natural images drawn from the COCO dataset. We find that our model replicates a range of findings from neuroscience, including attention-invariant tuning, inhibition of return, and attention-mediated scaling of activity. Understanding object based attention is both computationally interesting and a key problem for computational neuroscience.

The world provides us with data of multiple modalities. Intuitively, models fusingdata from different modalities outperform unimodal models, since more informationis aggregated. Recently, joining the success of deep learning, there is an influentialline of work on deep multimodal learning, which has remarkable empirical resultson various applications. However, theoretical justifications in this field are notablylacking.Can multimodal provably perform better than unimodal? In this paper, we answer this question under a most popular multimodal learningframework, which firstly encodes features from different modalities into a commonlatent space and seamlessly maps the latent representations into the task space. Weprove that learning with multiple modalities achieves a smaller population risk thanonly using its subset of modalities. The main intuition is that the former has moreaccurate estimate of the latent space representation. To the best of our knowledge,this is the first theoretical treatment to capture important qualitative phenomenaobserved in real multimodal applications. Combining with experiment results, weshow that multimodal learning does possess an appealing formal guarantee.

Deep neural networks are powerful machines for visual pattern recognition, but reasoning tasks that are easy for humans may still be difficult for neural models. Humans possess the ability to extrapolate reasoning strategies learned on simple problems to solve harder examples, often by thinking for longer. For example, a person who has learned to solve small mazes can easily extend the very same search techniques to solve much larger mazes by spending more time. In computers, this behavior is often achieved through the use of algorithms, which scale to arbitrarily hard problem instances at the cost of more computation. In contrast, the sequential computing budget of feed-forward neural networks is limited by their depth, and networks trained on simple problems have no way of extending their reasoning to accommodate harder problems. In this work, we show that recurrent networks trained to solve simple problems with few recurrent steps can indeed solve much more complex problems simply by performing additional recurrences during inference. We demonstrate this algorithmic behavior of recurrent networks on prefix sum computation, mazes, and chess. In all three domains, networks trained on simple problem instances are able to extend their reasoning abilities at test time simply by "thinking for longer."