Designing digital filters with desired phase responses at selected frequencies is essential in several communication and control applications. While phase compensation is generally achieved using all-pass filters, the all-pass constraint limits the range of realizable stable, causal filters that match the desired phase characteristics. In this paper, we present a method for designing real, stable discrete-time filters that match phase constraints at selected frequencies. Our approach enables control over the phase response without affecting stability and with low complexity. Specifically, we construct an nth-order filter to match phase constraints at n distinct frequencies (for even n), achieving exact phase interpolation with a low filter order. The derived methodology cascades suitably designed second order sections, referred to as filter blocks, and uses a fixed-point iteration scheme to meet the phase specifications, with tunable design parameters that enable explicit control over the margin of stability. Simulations confirm the utility of our method for obtaining stable filters that satisfy phase specifications even where comparable methods, such as all-pass designs, fail.

Diagnosing and mitigating changes in model fairness under distribution shift is an important component of the safe deployment of machine learning in healthcare settings. Importantly, the success of any mitigation strategy strongly depends on the \textit{structure} of the shift. Despite this, there has been little discussion of how to empirically assess the structure of a distribution shift that one is encountering in practice. In this work, we adopt a causal framing to motivate conditional independence tests as a key tool for characterizing distribution shifts. Using our approach in two medical applications, we show that this knowledge can help diagnose failures of fairness transfer, including cases where real-world shifts are more complex than is often assumed in the literature. Based on these results, we discuss potential remedies at each step of the machine learning pipeline.

Machine learning (ML) systems often exhibit unexpectedly poor behavior when they are deployed in real-world domains. We identify underspecification in ML pipelines as a key reason for these failures. An ML pipeline is the full procedure followed to train and validate a predictor. Such a pipeline is underspecified when it can return many distinct predictors with equivalently strong test performance. Underspecification is common in modern ML pipelines that primarily validate predictors on held-out data that follow the same distribution as the training data. Predictors returned by underspecified pipelines are often treated as equivalent based on their training domain performance, but we show here that such predictors can behave very differently in deployment domains. This ambiguity can lead to instability and poor model behavior in practice, and is a distinct failure mode from previously identified issues arising from structural mismatch between training and deployment domains. We provide evidence that underspecfication has substantive implications for practical ML pipelines, using examples from computer vision, medical imaging, natural language processing, clinical risk prediction based on electronic health records, and medical genomics. Our results show the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain.

Self-supervised pretraining followed by supervised fine-tuning has seen success in image recognition, especially when labeled examples are scarce, but has received limited attention in medical image analysis. This paper studies the effectiveness of self-supervised learning as a pre-training strategy for medical image classification. We conduct experiments on two distinct tasks: dermatology condition classification from digital camera images and multi-label chest X-ray classification, and demonstrate that self-supervised learning on ImageNet, followed by additional self-supervised learning on unlabeled domain-specific medical images significantly improves the accuracy of medical image classifiers. We introduce a novel Multi-Instance Contrastive Learning (MICLe) method that uses multiple images of the underlying pathology per patient case, when available, to construct more informative positive pairs for self-supervised learning. Combining our contributions, we achieve an improvement of 6.7% in top-1 accuracy and an improvement of 1.1% in mean AUC on dermatology and chest X-ray classification respectively, outperforming strong supervised baselines pretrained on ImageNet. In addition, we show that big self-supervised models are robust to distribution shift and can learn efficiently with a small number of labeled medical images.

Studies have shown that a dominant class of questions asked by visually impaired users on images of their surroundings involves reading text in the image. But today’s VQA models can not read! Our paper takes a first step towards addressing this problem. First, we introduce a new “TextVQA” dataset to facilitate progress on this important problem. Existing datasets either have a small proportion of questions about text (e.g., the VQA dataset) or are too small (e.g., the VizWiz dataset). TextVQA contains 45,336 questions on 28,408 images that require reasoning about text to answer. Second, we introduce a novel model architecture that reads text in the image, reasons about it in the context of the image and the question, and predicts an answer which might be a deduction based on the text and the image or composed of the strings found in the image. Consequently, we call our approach Look, Read, Reason & Answer (LoRRA). We show that LoRRA outperforms existing state-of-the-art VQA models on our TextVQA dataset. We find that the gap between human performance and machine performance is significantly larger on TextVQA than on VQA 2.0, suggesting that TextVQA is well-suited to benchmark progress along directions complementary to VQA 2.0.

Cooperation between nodes is essential for the functionality of a mobile ad-hoc network (MANET). However, since the nodes in a MANET are generally resource limited, some nodes could refuse service to other nodes to conserve their resources, thereby exhibiting selfish behavior. Also, since a MANET is often deployed in uncontrolled environments, some nodes could be compromised by an adversary and directed to act maliciously. A trust management framework in a MANETis useful to infer if nodes behave in a selfish or malicious manner, so that appropriate action could be taken, in order to maximize network performance. In this paper, we propose a scalable trust management scheme to partition and store an information network of trust metadata of nodes in a MANET. The simplicity of our scheme for trust metadata propagation and retrieval and its robustness to node failures, membership changes and mobility, make it a promising choice for trust management in a MANET. Simulation results that evaluate our scheme based on the trust management metrics we have defined demonstrate its performance benefits.

In a Delay-Tolerant Network (DTN), data originating from a source node may be delivered to the destination node, despite the non-existence of end-to-end connectivity between them at all times. In an adversarial environment such as a battlefield, DTN nodes could be compromised to launch Denial-of-Service (DoS) attacks by generating excess data, to cause an overflow of the limited resources of the legitimate nodes, hence decreasing the network throughput. A node may also display selfish behavior by generating more data than allowed, to increase its throughput and to decrease the latency of its data packets. In this paper, we term such a DoS attack and selfish data generation behavior, a resource-misuse attack. We study two types of resource-misuse attacks, breadth attacks and depth attacks. Accordingly, we propose different schemes to detect these attacks. Trace-driven simulations using both a synthetic and a real-world trace show that our detection schemes have low average detection latency and additionally, probabilistic detection of the depth attack has low false positive and false negative rates.