Myocardial infarction is the leading cause of death worldwide. In this paper, we design domain-inspired neural network models to detect myocardial infarction. First, we study the contribution of various leads. This systematic analysis, first of its kind in the literature, indicates that out of 15 ECG leads, data from the v6, vz, and ii leads are critical to correctly identify myocardial infarction. Second, we use this finding and adapt the ConvNetQuake neural network model--originally designed to identify earthquakes--to attain state-of-the-art classification results for myocardial infarction, achieving $99.43\%$ classification accuracy on a record-wise split, and $97.83\%$ classification accuracy on a patient-wise split. These two results represent cardiologist-level performance level for myocardial infarction detection after feeding only 10 seconds of raw ECG data into our model. Third, we show that our multi-ECG-channel neural network achieves cardiologist-level performance without the need of any kind of manual feature extraction or data pre-processing.
Aspect-based sentiment analysis (ABSA) task is a multi-grained task of natural language processing and consists of two subtasks: aspect term extraction (ATE) and aspect polarity classification (APC). Most of the existing work focuses on the subtask of aspect term polarity inferring and ignores the significance of aspect term extraction. Besides, the existing researches do not pay attention to the research of the Chinese-oriented ABSA task. Based on the local context focus (LCF) mechanism, this paper firstly proposes a multi-task learning model for Chinese-oriented aspect-based sentiment analysis, namely LCF-ATEPC. Compared with existing models, this model equips the capability of extracting aspect term and inferring aspect term polarity synchronously, moreover, this model is effective to analyze both Chinese and English comments simultaneously and the experiment on a multilingual mixed dataset proved its availability. By integrating the domain-adapted BERT model, the LCF-ATEPC model achieved the state-of-the-art performance of aspect term extraction and aspect polarity classification in four Chinese review datasets. Besides, the experimental results on the most commonly used SemEval-2014 task4 Restaurant and Laptop datasets outperform the state-of-the-art performance on the ATE and APC subtask.
An image preprocessing methodology based on Fourier analysis together with the Laguerre-Gauss Spatial Filter is proposed. This is an alternative to obtain features from aerial images that reduces the feature space significantly, preserving enough information for classification tasks. Experiments on a challenging data set of aerial images show that it is possible to learn a robust classifier from this transformed and smaller feature space using simple models, with similar performance to the complete feature space and more complex models.
Zero-shot learning, the task of learning to recognize new classes not seen during training, has received considerable attention in the case of 2D image classification. However despite the increasing ubiquity of 3D sensors, the corresponding 3D point cloud classification problem has not been meaningfully explored and introduces new challenges. This paper extends, for the first time, transductive Zero-Shot Learning (ZSL) and Generalized Zero-Shot Learning (GZSL) approaches to the domain of 3D point cloud classification. To this end, a novel triplet loss is developed that takes advantage of unlabeled test data. While designed for the task of 3D point cloud classification, the method is also shown to be applicable to the more common use-case of 2D image classification. An extensive set of experiments is carried out, establishing state-of-the-art for ZSL and GZSL in the 3D point cloud domain, as well as demonstrating the applicability of the approach to the image domain.
Toxic content detection aims to identify content that can offend or harm its recipients. Automated classifiers of toxic content need to be robust against adversaries who deliberately try to bypass filters. We propose a method of generating realistic model-agnostic attacks using a lexicon of toxic tokens, which attempts to mislead toxicity classifiers by diluting the toxicity signal either by obfuscating toxic tokens through character-level perturbations, or by injecting non-toxic distractor tokens. We show that these realistic attacks reduce the detection recall of state-of-the-art neural toxicity detectors, including those using ELMo and BERT, by more than 50% in some cases. We explore two approaches for defending against such attacks. First, we examine the effect of training on synthetically noised data. Second, we propose the Contextual Denoising Autoencoder (CDAE): a method for learning robust representations that uses character-level and contextual information to denoise perturbed tokens. We show that the two approaches are complementary, improving robustness to both character-level perturbations and distractors, recovering a considerable portion of the lost accuracy. Finally, we analyze the robustness characteristics of the most competitive methods and outline practical considerations for improving toxicity detectors.
Due to the simpleness and high efficiency, single-stage object detectors have been widely applied in many computer vision applications . However, the low correlation between the classification score and localization accuracy of the predicted detections has severely hurt the localization accuracy of models. In this paper, IoU-aware single-stage object detector is proposed to solve this problem. Specifically, IoU-aware single-stage object detector predicts the IoU for each detected box. Then the classification score and predicted IoU are multiplied to compute the final detection confidence, which is more correlated with the localization accuracy. The detection confidence is then used as the input of the subsequent NMS and COCO AP computation, which will substantially improve the localization accuracy of models. Sufficient experiments on COCO and PASCAL VOC datasets demonstrate the effectiveness of IoU-aware single-stage object detector on improving model's localization accuracy. Without whistles and bells, the proposed method can substantially improve AP by $1.7\%\sim1.9\%$ and AP75 by $2.2\%\sim2.5\%$ on COCO \textit{test-dev}. On PASCAL VOC, the proposed method can substantially improve AP by $2.9\%\sim4.4\%$ and AP80, AP90 by $4.6\%\sim10.2\%$. Code is available here: {https://github.com/ShengkaiWu/IoU-aware-single-stage-object-detector}.
Models with transparent inner structure and high classification performance are required to reduce potential risk and provide trust for users in domains like health care, finance, security, etc. However, existing models are hard to simultaneously satisfy the above two properties. In this paper, we propose a new hierarchical rule-based model for classification tasks, named Concept Rule Sets (CRS), which has both a strong expressive ability and a transparent inner structure. To address the challenge of efficiently learning the non-differentiable CRS model, we propose a novel neural network architecture, Multilayer Logical Perceptron (MLLP), which is a continuous version of CRS. Using MLLP and the Random Binarization (RB) method we proposed, we can search the discrete solution of CRS in continuous space using gradient descent and ensure the discrete CRS acts almost the same as the corresponding continuous MLLP. Experiments on 12 public data sets show that CRS outperforms the state-of-the-art approaches and the complexity of the learned CRS is close to the simple decision tree. Source code is available at https://github.com/12wang3/mllp.
This letter introduces a deep learning (DL) framework for direction-of-arrival (DOA) estimation. Previous works in DL context mostly consider a single or two target scenario which is a strong limitation in practice. Hence, in this work, we propose a DL framework for multiple signal classification (DeepMUSIC). We design multiple deep convolutional neural networks (CNNs), each of which is dedicated to a subregion of the angular spectrum. In particular, each CNN is fed with the array covariance matrix and it learns the MUSIC spectra of the corresponding angular subregion. We have shown, through simulations, that the proposed DeepMUSIC framework has superior estimation accuracy and exhibits less computational complexity in comparison with both DL and non-DL based techniques.
State-of-the-art object detectors rely on regressing and classifying an extensive list of possible anchors, which are divided into positive and negative samples based on their intersection-over-union (IoU) with corresponding groundtruth objects. Such a harsh split conditioned on IoU results in binary labels that are potentially noisy and challenging for training. In this paper, we propose to mitigate noise incurred by imperfect label assignment such that the contributions of anchors are dynamically determined by a carefully constructed cleanliness score associated with each anchor. Exploring outputs from both regression and classification branches, the cleanliness scores, estimated without incurring any additional computational overhead, are used not only as soft labels to supervise the training of the classification branch but also sample re-weighting factors for improved localization and classification accuracy. We conduct extensive experiments on COCO, and demonstrate, among other things, the proposed approach steadily improves RetinaNet by ~2% with various backbones.
Despite their numerous successes, there are many scenarios where adversarial risk metrics do not provide an appropriate measure of robustness. For example, test-time perturbations may occur in a probabilistic manner rather than being generated by an explicit adversary, while the poor train--test generalization of adversarial metrics can limit their usage to simple problems. Motivated by this, we develop a probabilistic robust risk framework, the statistically robust risk (SRR), which considers pointwise corruption distributions, as opposed to worst-case adversaries. The SRR provides a distinct and complementary measure of robust performance, compared to natural and adversarial risk. We show that the SRR admits estimation and training schemes which are as simple and efficient as for the natural risk: these simply require noising the inputs, but with a principled derivation for exactly how and why this should be done. Furthermore, we demonstrate both theoretically and experimentally that it can provide superior generalization performance compared with adversarial risks, enabling application to high-dimensional datasets.
Human activities often occur in specific scene contexts, e.g., playing basketball on a basketball court. Training a model using existing video datasets thus inevitably captures and leverages such bias (instead of using the actual discriminative cues). The learned representation may not generalize well to new action classes or different tasks. In this paper, we propose to mitigate scene bias for video representation learning. Specifically, we augment the standard cross-entropy loss for action classification with 1) an adversarial loss for scene types and 2) a human mask confusion loss for videos where the human actors are masked out. These two losses encourage learning representations that are unable to predict the scene types and the correct actions when there is no evidence. We validate the effectiveness of our method by transferring our pre-trained model to three different tasks, including action classification, temporal localization, and spatio-temporal action detection. Our results show consistent improvement over the baseline model without debiasing.
A key requirement for supervised machine learning is labeled training data, which is created by annotating unlabeled data with the appropriate class. Because this process can in many cases not be done by machines, labeling needs to be performed by human domain experts. This process tends to be expensive both in time and money, and is prone to errors. Additionally, reviewing an entire labeled dataset manually is often prohibitively costly, so many real world datasets contain mislabeled instances. To address this issue, we present in this paper a non-parametric end-to-end pipeline to find mislabeled instances in numerical, image and natural language datasets. We evaluate our system quantitatively by adding a small number of label noise to 29 datasets, and show that we find mislabeled instances with an average precision of more than 0.84 when reviewing our system's top 1\% recommendation. We then apply our system to publicly available datasets and find mislabeled instances in CIFAR-100, Fashion-MNIST, and others. Finally, we publish the code and an applicable implementation of our approach.
Few-shot learning is a fundamental task in computer vision that carries the promise of alleviating the need for exhaustively labeled data. Most few-shot learning approaches to date have focused on progressively more complex neural feature extractors and classifier adaptation strategies, as well as the refinement of the task definition itself. In this paper, we explore the hypothesis that a simple class-covariance-based distance metric, namely the Mahalanobis distance, adopted into a state of the art few-shot learning approach (CNAPS) can, in and of itself, lead to a significant performance improvement. We also discover that it is possible to learn adaptive feature extractors that allow useful estimation of the high dimensional feature covariances required by this metric from surprisingly few samples. The result of our work is a new "Simple CNAPS" architecture which has up to 9.2% fewer trainable parameters than CNAPS and performs up to 6.1% better than state of the art on the standard few-shot image classification benchmark dataset.
Many recent methods of zero-shot learning (ZSL) attempt to utilize generative model to generate the unseen visual samples from semantic descriptions and random noise. Therefore, the ZSL problem becomes a traditional supervised classification problem. However, most of the existing methods based on the generative model only focus on the quality of synthesized samples at the training stage, and ignore the importance of the zero-shot recognition stage. In this paper, we consider both the above two points and propose a novel approach. Specially, we select the Generative Adversarial Network (GAN) as our generative model. In order to improve the quality of synthesized samples, considering the internal relation of the semantic description in the semantic space as well as the fact that the seen and unseen visual information belong to different domains, we propose a bi-semantic reconstructing (BSR) component which contain two different semantic reconstructing regressors to lead the training of GAN. Since the semantic descriptions are available during the training stage, to further improve the ability of classifier, we combine the visual samples and semantic descriptions to train a classifier. At the recognition stage, we naturally utilize the BSR component to transfer the visual features and semantic descriptions, and concatenate them for classification. Experimental results show that our method outperforms the state of the art on several ZSL benchmark datasets with significant improvements.
Microblogs are widely used to express people's opinions and feelings in daily life. Sentiment analysis (SA) can timely detect personal sentiment polarities through analyzing text. Deep learning approaches have been broadly used in SA but still have not fully exploited syntax information. In this paper, we propose a syntax-based graph convolution network (GCN) model to enhance the understanding of diverse grammatical structures of Chinese microblogs. In addition, a pooling method based on percentile is proposed to improve the accuracy of the model. In experiments, for Chinese microblogs emotion classification categories including happiness, sadness, like, anger, disgust, fear, and surprise, the F-measure of our model reaches 82.32% and exceeds the state-of-the-art algorithm by 5.90%. The experimental results show that our model can effectively utilize the information of dependency parsing to improve the performance of emotion detection. What is more, we annotate a new dataset for Chinese emotion classification, which is open to other researchers.
In the last years, automatic classification of variable stars has received substantial attention. Using machine learning techniques for this task has proven to be quite useful. Typically, machine learning classifiers used for this task require to have a fixed training set, and the training process is performed offline. Upcoming surveys such as the Large Synoptic Survey Telescope (LSST) will generate new observations daily, where an automatic classification system able to create alerts online will be mandatory. A system with those characteristics must be able to update itself incrementally. Unfortunately, after training, most machine learning classifiers do not support the inclusion of new observations in light curves, they need to re-train from scratch. Naively re-training from scratch is not an option in streaming settings, mainly because of the expensive pre-processing routines required to obtain a vector representation of light curves (features) each time we include new observations. In this work, we propose a streaming probabilistic classification model; it uses a set of newly designed features that work incrementally. With this model, we can have a machine learning classifier that updates itself in real time with new observations. To test our approach, we simulate a streaming scenario with light curves from CoRot, OGLE and MACHO catalogs. Results show that our model achieves high classification performance, staying an order of magnitude faster than traditional classification approaches.
This work aims to examine one of the cornerstone problems of Musical Instrument Retrieval (MIR), in particular, instrument classification. IRMAS (Instrument recognition in Musical Audio Signals) data set is chosen for this purpose. The data includes musical clips recorded from various sources in the last century, thus having a wide variety of audio quality. We have presented a very concise summary of past work in this domain. Having implemented various supervised learning algorithms for this classification task, SVM classifier has outperformed the other state-of-the-art models with an accuracy of 79%. We also implemented Unsupervised techniques out of which Hierarchical Clustering has performed well.
Spoken Language Understanding (SLU) systems consist of several machine learning components operating together (e.g. intent classification, named entity recognition and resolution). Deep learning models have obtained state of the art results on several of these tasks, largely attributed to their better modeling capacity. However, an increase in modeling capacity comes with added costs of higher latency and energy usage, particularly when operating on low complexity devices. To address the latency and computational complexity issues, we explore a BranchyNet scheme on an intent classification scheme within SLU systems. The BranchyNet scheme when applied to a high complexity model, adds exit points at various stages in the model allowing early decision making for a set of queries to the SLU model. We conduct experiments on the Facebook Semantic Parsing dataset with two candidate model architectures for intent classification. Our experiments show that the BranchyNet scheme provides gains in terms of computational complexity without compromising model accuracy. We also conduct analytical studies regarding the improvements in the computational cost, distribution of utterances that egress from various exit points and the impact of adding more complexity to models with the BranchyNet scheme.
Early advancements in convolutional neural networks (CNNs) architectures are primarily driven by human expertise and by elaborate design processes. Recently, neural architecture search was proposed with the aim of automating the network design process and generating task-dependent architectures. While existing approaches have achieved competitive performance in image classification, they are not well suited to problems where the computational budget is limited for two reasons: (1) the obtained architectures are either solely optimized for classification performance, or only for one deployment scenario; (2) the search process requires vast computational resources in most approaches. To overcome these limitations, we propose an evolutionary algorithm for searching neural architectures under multiple objectives, such as classification performance and floating-point operations (FLOPs). The proposed method addresses the first shortcoming by populating a set of architectures to approximate the entire Pareto frontier through genetic operations that recombine and modify architectural components progressively. Our approach improves computational efficiency by carefully down-scaling the architectures during the search as well as reinforcing the patterns commonly shared among past successful architectures through Bayesian model learning. The integration of these two main contributions allows an efficient design of architectures that are competitive and in most cases outperform both manually and automatically designed architectures on benchmark image classification datasets: CIFAR, ImageNet, and human chest X-ray. The flexibility provided from simultaneously obtaining multiple architecture choices for different compute requirements further differentiates our approach from other methods in the literature. Code is available at https://github.com/mikelzc1990/nsganetv1
Convolutional neural networks (CNNs) have achieved state-of-the-art results on many visual recognition tasks. However, current CNN models still exhibit a poor ability to be invariant to spatial transformations of images. Intuitively, with sufficient layers and parameters, hierarchical combinations of convolution (matrix multiplication and non-linear activation) and pooling operations should be able to learn a robust mapping from transformed input images to transform-invariant representations. In this paper, we propose randomly transforming (rotation, scale, and translation) feature maps of CNNs during the training stage. This prevents complex dependencies of specific rotation, scale, and translation levels of training images in CNN models. Rather, each convolutional kernel learns to detect a feature that is generally helpful for producing the transform-invariant answer given the combinatorially large variety of transform levels of its input feature maps. In this way, we do not require any extra training supervision or modification to the optimization process and training images. We show that random transformation provides significant improvements of CNNs on many benchmark tasks, including small-scale image recognition, large-scale image recognition, and image retrieval. The code is available at https://github.com/jasonustc/caffe-multigpu/tree/TICNN.