Recent works show that the data distribution in a network's latent space is useful for estimating classification uncertainty and detecting Out-of-distribution (OOD) samples. To obtain a well-regularized latent space that is conducive for uncertainty estimation, existing methods bring in significant changes to model architectures and training procedures. In this paper, we present a lightweight, fast, and high-performance regularization method for Mahalanobis distance-based uncertainty prediction, and that requires minimal changes to the network's architecture. To derive Gaussian latent representation favourable for Mahalanobis Distance calculation, we introduce a self-supervised representation learning method that separates in-class representations into multiple Gaussians. Classes with non-Gaussian representations are automatically identified and dynamically clustered into multiple new classes that are approximately Gaussian. Evaluation on standard OOD benchmarks shows that our method achieves state-of-the-art results on OOD detection with minimal inference time, and is very competitive on predictive probability calibration. Finally, we show the applicability of our method to a real-life computer vision use case on microorganism classification.
相關內容
Hybrid RL is the setting where an RL agent has access to both offline data and online data by interacting with the real-world environment. In this work, we propose a new hybrid RL algorithm that combines an on-policy actor-critic method with offline data. On-policy methods such as policy gradient and natural policy gradient (NPG) have shown to be more robust to model misspecification, though sometimes it may not be as sample efficient as methods that rely on off-policy learning. On the other hand, offline methods that depend on off-policy training often require strong assumptions in theory and are less stable to train in practice. Our new approach integrates a procedure of off-policy training on the offline data into an on-policy NPG framework. We show that our approach, in theory, can obtain a best-of-both-worlds type of result -- it achieves the state-of-art theoretical guarantees of offline RL when offline RL-specific assumptions hold, while at the same time maintaining the theoretical guarantees of on-policy NPG regardless of the offline RL assumptions' validity. Experimentally, in challenging rich-observation environments, we show that our approach outperforms a state-of-the-art hybrid RL baseline which only relies on off-policy policy optimization, demonstrating the empirical benefit of combining on-policy and off-policy learning. Our code is publicly available at //github.com/YifeiZhou02/HNPG.
We consider a system model with two sources, a reliable source and an unreliable source, who are responsible for disseminating updates regarding a process to an age-based gossip network of $n$ nodes. Nodes wish to have fresh information, however, they have preference for packets that originated at the reliable source and are willing to sacrifice their version age of information by up to $G$ versions to switch from an unreliable packet to a reliable packet. We study how this protocol impacts the prevalence of unreliable packets at nodes in the network and their version age. Using a stochastic hybrid system (SHS) framework, we formulate analytical equations to characterize two quantities: expected fraction of nodes with unreliable packets and expected version age of information at network nodes. We show that as $G$ increases, fewer nodes have unreliable packet, however, their version age increases as well, thereby inducing a freshness-reliability trade-off in the network. We present numerical results to support our findings.
Irregular multivariate time series data is prevalent in the clinical and healthcare domains. It is characterized by time-wise and feature-wise irregularities, making it challenging for machine learning methods to work with. To solve this, we introduce a new model architecture composed of two modules: (1) DLA, a Dynamic Local Attention mechanism that uses learnable queries and feature-specific local windows when computing the self-attention operation. This results in aggregating irregular time steps raw input within each window to a harmonized regular latent space representation while taking into account the different features' sampling rates. (2) A hierarchical MLP mixer that processes the output of DLA through multi-scale patching to leverage information at various scales for the downstream tasks. Our approach outperforms state-of-the-art methods on three real-world datasets, including the latest clinical MIMIC IV dataset.
Domain fronting is a network communication technique that involves leveraging (or abusing) content delivery networks (CDNs) to disguise the final destination of network packets by presenting them as if they were intended for a different domain than their actual endpoint. This technique can be used for both benign and malicious purposes, such as circumventing censorship or hiding malware-related communications from network security systems. Since domain fronting has been known for a few years, some popular CDN providers have implemented traffic filtering approaches to curb its use at their CDN infrastructure. However, it remains unclear to what extent domain fronting has been mitigated. To better understand whether domain fronting can still be effectively used, we propose a systematic approach to discover CDNs that are still prone to domain fronting. To this end, we leverage passive and active DNS traffic analysis to pinpoint domain names served by CDNs and build an automated tool that can be used to discover CDNs that allow domain fronting in their infrastructure. Our results reveal that domain fronting is feasible in 22 out of 30 CDNs that we tested, including some major CDN providers like Akamai and Fastly. This indicates that domain fronting remains widely available and can be easily abused for malicious purposes.
Advances in lightweight neural networks have revolutionized computer vision in a broad range of IoT applications, encompassing remote monitoring and process automation. However, the detection of small objects, which is crucial for many of these applications, remains an underexplored area in current computer vision research, particularly for embedded devices. To address this gap, the paper proposes a novel adaptive tiling method that can be used on top of any existing object detector including the popular FOMO network for object detection on microcontrollers. Our experimental results show that the proposed tiling method can boost the F1-score by up to 225% while reducing the average object count error by up to 76%. Furthermore, the findings of this work suggest that using a soft F1 loss over the popular binary cross-entropy loss can significantly reduce the negative impact of imbalanced data. Finally, we validate our approach by conducting experiments on the Sony Spresense microcontroller, showcasing the proposed method's ability to strike a balance between detection performance, low latency, and minimal memory consumption.
Quantifying the effect of uncertainties in systems where only point evaluations in the stochastic domain but no regularity conditions are available is limited to sampling-based techniques. This work presents an adaptive sequential stratification estimation method that uses Latin Hypercube Sampling within each stratum. The adaptation is achieved through a sequential hierarchical refinement of the stratification, guided by previous estimators using local (i.e., stratum-dependent) variability indicators based on generalized polynomial chaos expansions and Sobol decompositions. For a given total number of samples $N$, the corresponding hierarchically constructed sequence of Stratified Sampling estimators combined with Latin Hypercube sampling is adequately averaged to provide a final estimator with reduced variance. Numerical experiments illustrate the procedure's efficiency, indicating that it can offer a variance decay proportional to $N^{-2}$ in some cases.
Deep generative models have emerged as a promising approach in the medical image domain to address data scarcity. However, their use for sequential data like respiratory sounds is less explored. In this work, we propose a straightforward approach to augment imbalanced respiratory sound data using an audio diffusion model as a conditional neural vocoder. We also demonstrate a simple yet effective adversarial fine-tuning method to align features between the synthetic and real respiratory sound samples to improve respiratory sound classification performance. Our experimental results on the ICBHI dataset demonstrate that the proposed adversarial fine-tuning is effective, while only using the conventional augmentation method shows performance degradation. Moreover, our method outperforms the baseline by 2.24% on the ICBHI Score and improves the accuracy of the minority classes up to 26.58%. For the supplementary material, we provide the code at //github.com/kaen2891/adversarial_fine-tuning_using_generated_respiratory_sound.
Detecting unusual patterns in graph data is a crucial task in data mining. However, existing methods often face challenges in consistently achieving satisfactory performance and lack interpretability, which hinders our understanding of anomaly detection decisions. In this paper, we propose a novel approach to graph anomaly detection that leverages the power of interpretability to enhance performance. Specifically, our method extracts an attention map derived from gradients of graph neural networks, which serves as a basis for scoring anomalies. In addition, we conduct theoretical analysis using synthetic data to validate our method and gain insights into its decision-making process. To demonstrate the effectiveness of our method, we extensively evaluate our approach against state-of-the-art graph anomaly detection techniques. The results consistently demonstrate the superior performance of our method compared to the baselines.
A case-cohort design is a two-phase sampling design frequently used to analyze censored survival data in a cost-effective way, where a subcohort is usually selected using simple random sampling or stratified simple random sampling. In this paper, we propose an efficient sampling procedure based on balanced sampling when selecting a subcohort in a case-cohort design. A sample selected via a balanced sampling procedure automatically calibrates auxiliary variables. When fitting a Cox model, calibrating sampling weights has been shown to lead to more efficient estimators of the regression coefficients (Breslow et al., 2009a, b). The reduced variabilities over its counterpart with a simple random sampling are shown via extensive simulation experiments. The proposed design and estimation procedure are also illustrated with the well-known National Wilms Tumor Study dataset.
Object detection typically assumes that training and test data are drawn from an identical distribution, which, however, does not always hold in practice. Such a distribution mismatch will lead to a significant performance drop. In this work, we aim to improve the cross-domain robustness of object detection. We tackle the domain shift on two levels: 1) the image-level shift, such as image style, illumination, etc, and 2) the instance-level shift, such as object appearance, size, etc. We build our approach based on the recent state-of-the-art Faster R-CNN model, and design two domain adaptation components, on image level and instance level, to reduce the domain discrepancy. The two domain adaptation components are based on H-divergence theory, and are implemented by learning a domain classifier in adversarial training manner. The domain classifiers on different levels are further reinforced with a consistency regularization to learn a domain-invariant region proposal network (RPN) in the Faster R-CNN model. We evaluate our newly proposed approach using multiple datasets including Cityscapes, KITTI, SIM10K, etc. The results demonstrate the effectiveness of our proposed approach for robust object detection in various domain shift scenarios.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191