The Deep neural networks (DNNs) have achieved great success on a variety of computer vision tasks, however, they are highly vulnerable to adversarial attacks. To address this problem, we propose to improve the local smoothness of the representation space, by integrating a margin-based triplet embedding regularization term into the classification objective, so that the obtained model learns to resist adversarial examples. The regularization term consists of two steps optimizations which find potential perturbations and punish them by a large margin in an iterative way. Experimental results on MNIST, CASIA-WebFace, VGGFace2 and MS-Celeb-1M reveal that our approach increases the robustness of the network against both feature and label adversarial attacks in simple object classification and deep face recognition.
The exploding cost and time needed for data labeling and model training are bottlenecks for training DNN models on large datasets. Identifying smaller representative data samples with strategies like active learning can help mitigate such bottlenecks. Previous works on active learning in NLP identify the problem of sampling bias in the samples acquired by uncertainty-based querying and develop costly approaches to address it. Using a large empirical study, we demonstrate that active set selection using the posterior entropy of deep models like FastText.zip (FTZ) is robust to sampling biases and to various algorithmic choices (query size and strategies) unlike that suggested by traditional literature. We also show that FTZ based query strategy produces sample sets similar to those from more sophisticated approaches (e.g ensemble networks). Finally, we show the effectiveness of the selected samples by creating tiny high-quality datasets, and utilizing them for fast and cheap training of large models. Based on the above, we propose a simple baseline for deep active text classification that outperforms the state-of-the-art. We expect the presented work to be useful and informative for dataset compression and for problems involving active, semi-supervised or online learning scenarios. Code and models are available at: https://github.com/drimpossible/Sampling-Bias-Active-Learning
Music is a mysterious language that conveys feeling and thoughts via different tones and timbre. For better understanding of timbre in music, we chose music data of 6 representative instruments, analysed their timbre features and classified them. Instead of the current trend of Neural Network for black-box classification, our project is based on a combination of MFCC and LPC, and augmented with a 6-dimensional feature vector designed by ourselves from observation and attempts. In our white-box model, we observed significant patterns of sound that distinguish different timbres, and discovered some connection between objective data and subjective senses. With a totally 32-dimensional feature vector and a naive all-pairs SVM, we achieved improved classification accuracy compared to a single tool. We also attempted to analyze music pieces downloaded from the Internet, found out different performance on different instruments, explored the reasons and suggested possible ways to improve the performance.
Can a text classifier generalize well for datasets where the text length is different? For example, when short reviews are sentiment-labeled, can these transfer to predict the sentiment of long reviews (i.e., short to long transfer), or vice versa? While unsupervised transfer learning has been well-studied for cross domain/lingual transfer tasks, Cross Length Transfer (CLT) has not yet been explored. One reason is the assumption that length difference is trivially transferable in classification. We show that it is not, because short/long texts differ in context richness and word intensity. We devise new benchmark datasets from diverse domains and languages, and show that existing models from similar tasks cannot deal with the unique challenge of transferring across text lengths. We introduce a strong baseline model called BaggedCNN that treats long texts as bags containing short texts. We propose a state-of-the-art CLT model called Length Transfer Networks (LeTraNets) that introduces a two-way encoding scheme for short and long texts using multiple training mechanisms. We test our models and find that existing models perform worse than the BaggedCNN baseline, while LeTraNets outperforms all models.
Dietary supplements are used by a large portion of the population, but information on their pharmacologic interactions is incomplete. To address this challenge, we present SUPP.AI, an application for browsing evidence of supplement-drug interactions (SDIs) extracted from the biomedical literature. We train a model to automatically extract supplement information and identify such interactions from the scientific literature. To address the lack of labeled data for SDI identification, we use labels of the closely related task of identifying drug-drug interactions (DDIs) for supervision. We fine-tune the contextualized word representations of the RoBERTa language model using labeled DDI data, and apply the fine-tuned model to identify supplement interactions. We extract 195k evidence sentences from 22M articles (P=0.82, R=0.58, F1=0.68) for 60k interactions. We create the SUPP.AI application for users to search evidence sentences extracted by our model. SUPP.AI is an attempt to close the information gap on dietary supplements by making up-to-date evidence on SDIs more discoverable for researchers, clinicians, and consumers.
In this paper, we focus on the classification of books using short descriptive texts (cover blurbs) and additional metadata. Building upon BERT, a deep neural language model, we demonstrate how to combine text representations with metadata and knowledge graph embeddings, which encode author information. Compared to the standard BERT approach we achieve considerably better results for the classification task. For a more coarse-grained classification using eight labels we achieve an F1- score of 87.20, while a detailed classification using 343 labels yields an F1-score of 64.70. We make the source code and trained models of our experiments publicly available
Deep learning Convolutional Neural Network (CNN) models are powerful classification models but require a large amount of training data. In niche domains such as bird acoustics, it is expensive and difficult to obtain a large number of training samples. One method of classifying data with a limited number of training samples is to employ transfer learning. In this research, we evaluated the effectiveness of birdcall classification using transfer learning from a larger base dataset (2814 samples in 46 classes) to a smaller target dataset (351 samples in 10 classes) using the ResNet-50 CNN. We obtained 79% average validation accuracy on the target dataset in 5-fold cross-validation. The methodology of transfer learning from an ImageNet-trained CNN to a project-specific and a much smaller set of classes and images was extended to the domain of spectrogram images, where the base dataset effectively played the role of the ImageNet.
We report on an open-source implementation for distributed function minimization on top of Apache Spark by using gradient and quasi-Newton methods. We show-case it with an application to Optimal Transport and some scalability tests on classification and regression problems.
Gaussian process classification (GPC) provides a flexible and powerful statistical framework describing joint distributions over function space. Conventional GPCs however suffer from (i) poor scalability for big data due to the full kernel matrix, and (ii) intractable inference due to the non-Gaussian likelihoods. Hence, various scalable GPCs have been proposed through (i) the sparse approximation built upon a small inducing set to reduce the time complexity; and (ii) the approximate inference to derive analytical evidence lower bound (ELBO). However, these scalable GPCs equipped with analytical ELBO are limited to specific likelihoods or additional assumptions. In this work, we present a unifying framework which accommodates scalable GPCs using various likelihoods. Analogous to GP regression (GPR), we introduce additive noises to augment the probability space for (i) the GPCs with step, (multinomial) probit and logit likelihoods via the internal variables; and particularly, (ii) the GPC using softmax likelihood via the noise variables themselves. This leads to unified scalable GPCs with analytical ELBO by using variational inference. Empirically, our GPCs showcase better results than state-of-the-art scalable GPCs for extensive binary/multi-class classification tasks with up to two million data points.
The scarcity of labeled training data often prohibits the internationalization of NLP models to multiple languages. Recent developments in cross-lingual understanding (XLU) has made progress in this area, trying to bridge the language barrier using language universal representations. However, even if the language problem was resolved, models trained in one language would not transfer to another language perfectly due to the natural domain drift across languages and cultures. We consider the setting of semi-supervised cross-lingual understanding, where labeled data is available in a source language (English), but only unlabeled data is available in the target language. We combine state-of-the-art cross-lingual methods with recently proposed methods for weakly supervised learning such as unsupervised pre-training and unsupervised data augmentation to simultaneously close both the language gap and the domain gap in XLU. We show that addressing the domain gap is crucial. We improve over strong baselines and achieve a new state-of-the-art for cross-lingual document classification.
Over the past decade, convolutional neural networks (CNNs) have become the driving force of an ever-increasing set of applications, achieving state-of-the-art performance. Most of the modern CNN architectures are composed of many convolutional and fully connected layers and typically require thousands or millions of parameters to learn. CNNs have also been effective in the detection of Event-Related Potentials from electroencephalogram (EEG) signals, notably the P300 component which is frequently employed in Brain-Computer Interfaces (BCIs). However, for this task, the increase in detection rates compared to approaches based on human-engineered features has not been as impressive as in other areas and might not justify such a large number of parameters. In this paper, we study the performances of existing CNN architectures with diverse complexities for single-trial within-subject and cross-subject P300 detection on four different datasets. We also proposed SepConv1D, a very simple CNN architecture consisting of a single depthwise separable 1D convolutional layer followed by a fully connected Sigmoid classification neuron. We found that with as few as four filters in its convolutional layer and a small overall number of parameters, SepConv1D obtained competitive performances in the four datasets. We believe this may represent an important step towards building simpler, cheaper, faster, and more portable BCIs.
Large scale object detection datasets are constantly increasing their size in terms of the number of classes and annotations count. Yet, the number of object-level categories annotated in detection datasets is an order of magnitude smaller than image-level classification labels. State-of-the art object detection models are trained in a supervised fashion and this limits the number of object classes they can detect. In this paper, we propose a novel weight transfer network (WTN) to effectively and efficiently transfer knowledge from classification network's weights to detection network's weights to allow detection of novel classes without box supervision. We first introduce input and feature normalization schemes to curb the under-fitting during training of a vanilla WTN. We then propose autoencoder-WTN (AE-WTN) which uses reconstruction loss to preserve classification network's information over all classes in the target latent space to ensure generalization to novel classes. Compared to vanilla WTN, AE-WTN obtains absolute performance gains of 6% on two Open Images evaluation sets with 500 seen and 57 novel classes respectively, and 25% on a Visual Genome evaluation set with 200 novel classes. The code is available at https://github.com/xternalz/AE-WTN.
Prediction problems often admit competing models that perform almost equally well. This effect challenges key assumptions in machine learning when competing models assign conflicting predictions. In this paper, we define predictive multiplicity as the ability of a prediction problem to admit competing models with conflicting predictions. We introduce formal measures to evaluate the severity of predictive multiplicity and develop integer programming tools to compute them exactly for linear classification problems. We apply our tools to measure predictive multiplicity in recidivism prediction problems. Our results show that real-world datasets may admit competing models that assign wildly conflicting predictions, and motivate the need to measure and report predictive multiplicity in model development.
Zero-shot learning (ZSL) aims to recognize instances of unseen classes solely based on the semantic descriptions of the classes. Existing algorithms usually formulate it as a semantic-visual correspondence problem, by learning mappings from one feature space to the other. Despite being reasonable, previous approaches essentially discard the highly precious discriminative power of visual features in an implicit way, and thus produce undesirable results. We instead reformulate ZSL as a conditioned visual classification problem, i.e., classifying visual features based on the classifiers learned from the semantic descriptions. With this reformulation, we develop algorithms targeting various ZSL settings: For the conventional setting, we propose to train a deep neural network that directly generates visual feature classifiers from the semantic attributes with an episode-based training scheme; For the generalized setting, we concatenate the learned highly discriminative classifiers for seen classes and the generated classifiers for unseen classes to classify visual features of all classes; For the transductive setting, we exploit unlabeled data to effectively calibrate the classifier generator using a novel learning-without-forgetting self-training mechanism and guide the process by a robust generalized cross-entropy loss. Extensive experiments show that our proposed algorithms significantly outperform state-of-the-art methods by large margins on most benchmark datasets in all the ZSL settings. Our code is available at \url{https://github.com/kailigo/cvcZSL}
Out-of-domain (OOD) detection for low-resource text classification is a realistic but understudied task. The goal is to detect the OOD cases with limited in-domain (ID) training data, since we observe that training data is often insufficient in machine learning applications. In this work, we propose an OOD-resistant Prototypical Network to tackle this zero-shot OOD detection and few-shot ID classification task. Evaluation on real-world datasets show that the proposed solution outperforms state-of-the-art methods in zero-shot OOD detection task, while maintaining a competitive performance on ID classification task.
In recent years, Deep Neural Networks (DNNs) have had a dramatic impact on a variety of problems that were long considered very difficult, e. g., image classification and automatic language translation to name just a few. The accuracy of modern DNNs in classification tasks is remarkable indeed. At the same time, attackers have devised powerful methods to construct specially-crafted malicious inputs (often referred to as adversarial examples) that can trick DNNs into mis-classifying them. What is worse is that despite the many defense mechanisms proposed to protect DNNs against adversarial attacks, attackers are often able to circumvent these defenses, rendering them useless. This state of affairs is extremely worrying, especially since machine learning systems get adopted at scale. In this paper, we propose a scientific evaluation methodology aimed at assessing the quality, efficacy, robustness and efficiency of randomized defenses to protect DNNs against adversarial examples. Using this methodology, we evaluate a variety of defense mechanisms. In addition, we also propose a defense mechanism we call Randomly Perturbed Ensemble Neural Networks (RPENNs). We provide a thorough and comprehensive evaluation of the considered defense mechanisms against a white-box attacker model, six different adversarial attack methods and using the ILSVRC2012 validation data set.
This paper brings deep learning at the forefront of research into Time Series Classification (TSC). TSC is the area of machine learning tasked with the categorization (or labelling) of time series. The last few decades of work in this area have led to significant progress in the accuracy of classifiers, with the state of the art now represented by the HIVE-COTE algorithm. While extremely accurate, HIVE-COTE cannot be applied to many real-world datasets because of its high training time complexity in O(N2 * T4) for a dataset with N time series of length T. For example, it takes HIVE-COTE more than 8 days to learn from a small dataset with N = 1500 time series of short length T = 46. Meanwhile deep learning has received enormous attention because of its high accuracy and scalability. Recent approaches to deep learning for TSC have been scalable, but less accurate than HIVE-COTE. We introduce InceptionTime - an ensemble of deep Convolutional Neural Network (CNN) models, inspired by the Inception-v4 architecture. Our experiments show that InceptionTime is on par with HIVE-COTE in terms of accuracy while being much more scalable: not only can it learn from 1,500 time series in one hour but it can also learn from 8M time series in 13 hours, a quantity of data that is fully out of reach of HIVE-COTE.
To train accurate deep object detectors under the extreme foreground-background imbalance, heuristic sampling methods are always necessary, which either re-sample a subset of all training samples (hard sampling methods, \eg biased sampling, OHEM), or use all training samples but re-weight them discriminatively (soft sampling methods, \eg Focal Loss, GHM). In this paper, we challenge the necessity of such hard/soft sampling methods for training accurate deep object detectors. While previous studies have shown that training detectors without heuristic sampling methods would significantly degrade accuracy, we reveal that this degradation comes from an unreasonable classification gradient magnitude caused by the imbalance, rather than a lack of re-sampling/re-weighting. Motivated by our discovery, we propose a simple yet effective \emph{Sampling-Free} mechanism to achieve a reasonable classification gradient magnitude by initialization and loss scaling. Unlike heuristic sampling methods with multiple hyperparameters, our Sampling-Free mechanism is fully data diagnostic, without laborious hyperparameters searching. We verify the effectiveness of our method in training anchor-based and anchor-free object detectors, where our method always achieves higher detection accuracy than heuristic sampling methods on COCO and PASCAL VOC datasets. Our Sampling-Free mechanism provides a new perspective to address the foreground-background imbalance. Our code is released at \url{https://github.com/ChenJoya/sampling-free}.
The use of semantic segmentation for masking and cropping input images has proven to be a significant aid in medical imaging classification tasks by decreasing the noise and variance of the training dataset. However, implementing this approach with classical methods is challenging: the cost of obtaining a dense segmentation is high, and the precise input area that is most crucial to the classification task is difficult to determine a-priori. We propose a novel joint-training deep reinforcement learning framework for image augmentation. A segmentation network, weakly supervised with policy gradient optimization, acts as an agent, and outputs masks as actions given samples as states, with the goal of maximizing reward signals from the classification network. In this way, the segmentation network learns to mask unimportant imaging features. Our method, Adversarial Policy Gradient Augmentation (APGA), shows promising results on Stanford's MURA dataset and on a hip fracture classification task with an increase in global accuracy of up to 7.33% and improved performance over baseline methods in 9/10 tasks evaluated. We discuss the broad applicability of our joint training strategy to a variety of medical imaging tasks.
As a step toward better document-level understanding, we explore classification of a sequence of sentences into their corresponding categories, a task that requires understanding sentences in context of the document. Recent successful models for this task have used hierarchical models to contextualize sentence representations, and Conditional Random Fields (CRFs) to incorporate dependencies between subsequent labels. In this work, we show that pretrained language models, BERT (Devlin et al., 2018) in particular, can be used for this task to capture contextual dependencies without the need for hierarchical encoding nor a CRF. Specifically, we construct a joint sentence representation that allows BERT Transformer layers to directly utilize contextual information from all words in all sentences. Our approach achieves state-of-the-art results on four datasets, including a new dataset of structured scientific abstracts.