A family of loss functions built on pair-based computation have been proposed in the literature which provide a myriad of solutions for deep metric learning. In this paper, we provide a general weighting framework for understanding recent pair-based loss functions. Our contributions are three-fold: (1) we establish a General Pair Weighting (GPW) framework, which casts the sampling problem of deep metric learning into a unified view of pair weighting through gradient analysis, providing a powerful tool for understanding recent pair-based loss functions; (2) we show that with GPW, various existing pair-based methods can be compared and discussed comprehensively, with clear differences and key limitations identified; (3) we propose a new loss called multi-similarity loss (MS loss) under the GPW, which is implemented in two iterative steps (i.e., mining and weighting). This allows it to fully consider three similarities for pair weighting, providing a more principled approach for collecting and weighting informative pairs. Finally, the proposed MS loss obtains new state-of-the-art performance on four image retrieval benchmarks, where it outperforms the most recent approaches, such as ABE [14] and HTL [4], by a large margin, e.g., , and 80.9% → 88.0% on In-Shop Clothes Retrieval dataset at Recall@1. Code is available at https://github. com/MalongTech/research-ms-loss arXiv:1904.06627v3 [cs.CV]

We present a novel hierarchical triplet loss (HTL) capable of automatically collecting informative training samples (triplets) via a defined hierarchical tree that encodes global context information. This allows us to cope with the main limitation of random sampling in training a conventional triplet loss, which is a central issue for deep metric learning. Our main contributions are two-fold. (i) we construct a hierarchical class-level tree where neighboring classes are merged recursively. The hierarchical structure naturally captures the intrinsic data distribution over the whole dataset. (ii) we formulate the problem of triplet collection by introducing a new violate margin, which is computed dynamically based on the designed hierarchical tree. This allows it to automatically select meaningful hard samples with the guide of global context. It encourages the model to learn more discriminative features from visual similar classes, leading to faster convergence and better performance. Our method is evaluated on the tasks of image retrieval and face recognition, where it outperforms the standard triplet loss substantially by 1%18%. It achieves new state-of-the-art performance on a number of benchmarks, with much fewer learning iterations.

IMPORTANCE Mounting evidence suggests that sex differences exist in the pathologic trajectory of Alzheimer disease. Previous literature shows elevated levels of cerebrospinal fluid tau in women compared with men as a function of apolipoprotein E (APOE) ε4 status and β-amyloid (Aβ). What remains unclear is the association of sex with regional tau deposition in clinically normal individuals. OBJECTIVE To examine sex differences in the cross-sectional association between Aβ and regional tau deposition as measured with positron emission tomography (PET). DESIGN, SETTING AND PARTICIPANTS This is a study of 2 cross-sectional, convenience-sampled cohorts of clinically normal individuals who received tau and Aβ PET scans. Data were collected between January 2016 and February 2018 from 193 clinically normal individuals from the Harvard Aging Brain Study (age range, 55-92 years; 118 women [61%]) who underwent carbon 11-labeled Pittsburgh Compound B and flortaucipir F 18 PET and 103 clinically normal individuals from the Alzheimer's Disease Neuroimaging Initiative (age range, 63-94 years; 55 women [51%]) who underwent florbetapir and flortaucipir F 18 PET. MAIN OUTCOMES AND MEASURES A main association of sex with regional tau in the entorhinal cortices, inferior temporal lobe, and a meta-region of interest, which was a composite of regions in the temporal lobe. Associations between sex and global Aβ as well as sex and APOE ε4 on these regions after controlling for age were also examined. RESULTS The mean (SD) age of all individuals was 74.2 (7.6) years (81 APOE ε4 carriers [31%]; 89 individuals [30%] with high Aβ). There was no clear association of sex with regional tau that was replicated across studies. However, in both cohorts, clinically normal women exhibited higher entorhinal cortical tau than men (meta-analytic estimate: β [male] = −0.11 [0.05]; 95% CI, −0.21 to −0.02; P = .02), which was associated with individuals with higher Aβ burden. A sex by APOE ε4 interaction was not associated with regional tau (meta-analytic estimate: β [male, APOE ε4+] = −0.15 [0.09]; 95% CI, −0.32 to 0.01; P = .07). CONCLUSIONS AND RELEVANCE Early tau deposition was elevated in women compared with men in individuals on the Alzheimer disease trajectory. These findings lend support to a growing body of literature that highlights a biological underpinning for sex differences in Alzheimer disease risk.

We present a simple yet efficient approach capable of training deep neural networks on large-scale weakly-supervised web images, which are crawled raw from the Internet by using text queries, without any human annotation. We develop a principled learning strategy by leveraging curriculum learning, with the goal of handling a massive amount of noisy labels and data imbalance effectively. We design a new learning curriculum by measuring the complexity of data using its distribution density in a feature space, and rank the complexity in an unsupervised manner. This allows for an efficient implementation of curriculum learning on large-scale web images, resulting in a highperformance CNN model, where the negative impact of noisy labels is reduced substantially. Importantly, we show by experiments that those images with highly noisy labels can surprisingly improve the generalization capability of the model, by serving as a manner of regularization. Our approaches obtain state-of-the-art performance on four benchmarks: WebVision, ImageNet, Clothing-1M and Food-101. With an ensemble of multiple models, we achieved a top-5 error rate of 5.2% on the WebVision challenge [18] for 1000-category classification. This result was the top performance by a wide margin, outperforming second place by a nearly 50% relative error rate. Code and models are available at: https://github.com/MalongTech/CurriculumNet.