Batch Normalization (BN) (Ioffe and Szegedy 2015) normalizes the features of an input image via statistics of a batch of images and this batch information is considered as batch noise that will be brought to the features of an instance by BN. We offer a point of view that self-attention mechanism can help regulate the batch noise by enhancing instance-specific information. Based on this view, we propose combining BN with a self-attention mechanism to adjust the batch noise and give an attention-based version of BN called Instance Enhancement Batch Normalization (IEBN) which recalibrates channel information by a simple linear transformation. IEBN outperforms BN with a light parameter increment in various visual tasks universally for different network structures and benchmark data sets. Besides, even if under the attack of synthetic noise, IEBN can still stabilize network training with good generalization. The code of IEBN is available at https://github.com/gbup-group/IEBN

The HDoutliers algorithm is a powerful unsupervised algorithm for detecting anomalies in high-dimensional data, with a strong theoretical foundation. However, it suffers from some limitations that significantly hinder its performance level, under certain circumstances. In this article, we propose an algorithm that addresses these limitations. We define an anomaly as an observation that deviates markedly from the majority with a large distance gap. An approach based on extreme value theory is used for the anomalous threshold calculation. Using various synthetic and real datasets, we demonstrate the wide applicability and usefulness of our algorithm, which we call the stray algorithm. We also demonstrate how this algorithm can assist in detecting anomalies present in other data structures using feature engineering. We show the situations where the stray algorithm outperforms the HDoutliers algorithm both in accuracy and computational time. This framework is implemented in the open source R package stray.

Learning a good speaker embedding is important for many automatic speaker recognition tasks, including verification, identification and diarization. The embeddings learned by softmax are not discriminative enough for open-set verification tasks. Angular based embedding learning target can achieve such discriminativeness by optimizing angular distance and adding margin penalty. We apply several different popular angular margin embedding learning strategies in this work and explicitly compare their performance on Voxceleb speaker recognition dataset. Observing the fact that encouraging inter-class separability is important when applying angular based embedding learning, we propose an exclusive inter-class regularization as a complement for angular based loss. We verify the effectiveness of these methods for learning a discriminative embedding space on ASV task with several experiments. These methods together, we manage to achieve an impressive result with 16.5% improvement on equal error rate (EER) and 18.2% improvement on minimum detection cost function comparing with baseline softmax systems.

Dynamic ensembling of classifiers is an effective approach in processing label-imbalanced classifications. However, in dynamic ensemble methods, the combination of classifiers is usually determined by the local competence and conventional regularization methods are difficult to apply, leaving the technique prone to overfitting. In this paper, focusing on the binary label-imbalanced classification field, a novel method of Adaptive Ensemble of classifiers with Regularization (AER) has been proposed. The method deals with the overfitting problem from a perspective of implicit regularization. Specifically, it leverages the properties of Stochastic Gradient Descent (SGD) to obtain the solution with the minimum norm to achieve regularization, and interpolates ensemble weights via the global geometry of data to further prevent overfitting. The method enjoys a favorable time and memory complexity, and theoretical proofs show that algorithms implemented with AER paradigm have time and memory complexities upper-bounded by their original implementations. Furthermore, the proposed AER method is tested with a specific implementation based on Gradient Boosting Machine (XGBoost) on the three datasets: UCI Bioassay, KEEL Abalone19, and a set of GMM-sampled artificial dataset. Results show that the proposed AER algorithm can outperform the major existing algorithms based on multiple metrics, and Mcnemar's tests are applied to validate performance superiorities. To summarize, this work complements regularization for dynamic ensemble methods and develops an algorithm superior in grasping both the global and local geometry of data to alleviate overfitting in imbalanced data classification.

We demonstrate the existence of universal adversarial perturbations, which can fool a family of audio processing architectures, for both targeted and untargeted attacks. To the best of our knowledge, this is the first study on generating universal adversarial perturbations for audio processing systems. We propose two methods for finding such perturbations. The first method is based on an iterative, greedy approach that is well-known in computer vision: it aggregates small perturbations to the input so as to push it to the decision boundary. The second method, which is the main technical contribution of this work, is a novel penalty formulation, which finds targeted and untargeted universal adversarial perturbations. Differently from the greedy approach, the penalty method minimizes an appropriate objective function on a batch of samples. Therefore, it produces more successful attacks when the number of training samples is limited. Moreover, we provide a proof that the proposed penalty method theoretically converges to a solution that corresponds to universal adversarial perturbations. We report comprehensive experiments, showing attack success rates higher than 91.1% and 74.7% for targeted and untargeted attacks, respectively.

Weakly supervised object localization (WSOL) models aim to locate objects of interest in an image after being trained only on data with coarse image level labels. Deep learning models for WSOL rely typically on convolutional attention maps with no constraints on the regions of interest which allows them to select any region, making them vulnerable to false positive regions. This issue occurs in many application domains, e.g., medical image analysis, where interpretability is central to the prediction. In order to improve the localization reliability, we propose a deep learning framework for WSOL with pixel level localization. It is composed of two sequential sub-networks: a localizer that localizes regions of interest; followed by a classifier that classifies them. Within its end-to-end training, we incorporate the prior knowledge stating that in an agnostic-class setup an image is more likely to contain relevant --object of interest-- and irrelevant regions --noise--. Based on the conditional entropy (CE) measured at the classifier, the localizer is driven to spot relevant regions (low CE), and irrelevant regions (high CE). Our framework is able to recover large discriminative regions using our recursive erasing algorithm that we incorporate within the backpropagation during training. Moreover, the framework handles intrinsically multi-instances. Experimental results on public datasets with medical images (GlaS colon cancer) and natural images (Caltech-UCSD Birds-200-2011, Oxford flower 102) show that, compared to state of the art WSOL methods, our framework can provide significant improvements in terms of image-level classification, pixel-level localization, and robustness to overfitting when dealing with few training samples. A public reproducible PyTorch implementation is provided in: https://github.com/sbelharbi/wsol-min-max-entropy-interpretability .

Building a Massive Corpus for Named Entity Recognition using Free Open Data Sources Daniel Specht Menezes Departamento de Inform atica PUC-Rio Rio de Janeiro, Brazil danielssmenezes@gmail.com Abstract --With the recent progress in machine learning, boosted by techniques such as deep learning, many tasks can be successfully solved once a large enough dataset is available for training. Nonetheless, human-annotated datasets are often expensive to produce, especially when labels are fine-grained, as is the case of Named Entity Recognition (NER), a task that operates with labels on a word-level. In this paper, we propose a method to automatically generate labeled datasets for NER from public data sources by exploiting links and structured data from DBpedia and Wikipedia. Due to the massive size of these data sources, the resulting dataset - SESAME 1 - is composed of millions of labeled sentences. We detail the method to generate the dataset, report relevant statistics, and design a baseline using a neural network, showing that our dataset helps building better NER predictors.

Using class labels to represent class similarity is a typical approach to training deep hashing systems for retrieval; samples from the same or different classes take binary 1 or 0 similarity values. This similarity does not model the full rich knowledge of semantic relations that may be present between data points. In this work we build upon the idea of using semantic hierarchies to form distance metrics between all available sample labels; for example cat to dog has a smaller distance than cat to guitar. We combine this type of semantic distance into a loss function to promote similar distances between the deep neural network embeddings. We also introduce an empirical Kullback-Leibler divergence loss term to promote binarization and uniformity of the embeddings. We test the resulting SHREWD method and demonstrate improvements in hierarchical retrieval scores using compact, binary hash codes instead of real valued ones, and show that in a weakly supervised hashing setting we are able to learn competitively without explicitly relying on class labels, but instead on similarities between labels.

Multi-view data are becoming common in real-world modeling tasks and many multi-view data clustering algorithms have thus been proposed. The existing algorithms usually focus on the cooperation of different views in the original space but neglect the influence of the hidden information among these different visible views, or they only consider the hidden information between the views. The algorithms are therefore not efficient since the available information is not fully excavated, particularly the otherness information in different views and the consistency information between them. In practice, the otherness and consistency information in multi-view data are both very useful for effective clustering analyses. In this study, a Multi-View clustering algorithm developed with the Cooperation of Visible and Hidden views, i.e., MV-Co-VH, is proposed. The MV-Co-VH algorithm first projects the multiple views from different visible spaces to the common hidden space by using the non-negative matrix factorization (NMF) strategy to obtain the common hidden view data. Collaborative learning is then implemented in the clustering procedure based on the visible views and the shared hidden view. The results of extensive experiments on UCI multi-view datasets and real-world image multi-view datasets show that the clustering performance of the proposed algorithm is competitive with or even better than that of the existing algorithms.

Label manipulation attacks are a subclass of data poisoning attacks in adversarial machine learning used against different applications, such as malware detection. These types of attacks represent a serious threat to detection systems in environments having high noise rate or uncertainty, such as complex networks and Internet of Thing (IoT). Recent work in the literature has suggested using the $K$-Nearest Neighboring (KNN) algorithm to defend against such attacks. However, such an approach can suffer from low to wrong detection accuracy. In this paper, we design an architecture to tackle the Android malware detection problem in IoT systems. We develop an attack mechanism based on Silhouette clustering method, modified for mobile Android platforms. We proposed two Convolutional Neural Network (CNN)-type deep learning algorithms against this \emph{Silhouette Clustering-based Label Flipping Attack (SCLFA)}. We show the effectiveness of these two defense algorithms - \emph{Label-based Semi-supervised Defense (LSD)} and \emph{clustering-based Semi-supervised Defense (CSD)} - in correcting labels being attacked. We evaluate the performance of the proposed algorithms by varying the various machine learning parameters on three Android datasets: Drebin, Contagio, and Genome and three types of features: API, intent, and permission. Our evaluation shows that using random forest feature selection and varying ratios of features can result in an improvement of up to 19\% accuracy when compared with the state-of-the-art method in the literature.