Adversarial training (AT) as a regularization method has proved its effectiveness in various tasks, such as image classification and text classification. Though there are successful applications of AT in many tasks of natural language processing (NLP), the mechanism behind it is still unclear. In this paper, we aim to apply AT on machine reading comprehension (MRC) and study its effects from multiple perspectives. We experiment with three different kinds of RC tasks: span-based RC, span-based RC with unanswerable questions and multi-choice RC. The experimental results show that the proposed method can improve the performance significantly and universally on SQuAD1.1, SQuAD2.0 and RACE. With virtual adversarial training (VAT), we explore the possibility of improving the RC models with semi-supervised learning and prove that examples from a different task are also beneficial. We also find that AT helps little in defending against artificial adversarial examples, but AT helps the model to learn better on examples that contain more low-frequency words.
We study the potential for interaction in natural language classification. We add a limited form of interaction for intent classification, where users provide an initial query using natural language, and the system asks for additional information using binary or multi-choice questions. At each turn, our system decides between asking the most informative question or making the final classification prediction.The simplicity of the model allows for bootstrapping of the system without interaction data, instead relying on simple crowdsourcing tasks. We evaluate our approach on two domains, showing the benefit of interaction and the advantage of learning to balance between asking additional questions and making the final prediction.
Generating temporal action proposals remains a very challenging problem, where the main issue lies in predicting precise temporal proposal boundaries and reliable action confidence in long and untrimmed real-world videos. In this paper, we propose an efficient and unified framework to generate temporal action proposals named Dense Boundary Generator (DBG), which draws inspiration from boundary-sensitive methods and implements boundary classification and action completeness regression for densely distributed proposals. In particular, the DBG consists of two modules: Temporal boundary classification (TBC) and Action-aware completeness regression (ACR). The TBC aims to provide two temporal boundary confidence maps by low-level two-stream features, while the ACR is designed to generate an action completeness score map by high-level action-aware features. Moreover, we introduce a dual stream BaseNet (DSB) to encode RGB and optical flow information, which helps to capture discriminative boundary and actionness features. Extensive experiments on popular benchmarks ActivityNet-1.3 and THUMOS14 demonstrate the superiority of DBG over the state-of-the-art proposal generator (e.g., MGG and BMN). Our code will be made available upon publication.
We propose a novel framework to understand the text by converting sentences or articles into video-like 3-dimensional tensors. Each frame, corresponding to a slice of the tensor, is a word image that is rendered by the word's shape. The length of the tensor equals to the number of words in the sentence or article. The proposed transformation from the text to a 3-dimensional tensor makes it very convenient to implement an $n$-gram model with convolutional neural networks for text analysis. Concretely, we impose a 3-dimensional convolutional kernel on the 3-dimensional text tensor. The first two dimensions of the convolutional kernel size equal the size of the word image and the last dimension of the kernel size is $n$. That is, every time when we slide the 3-dimensional kernel over a word sequence, the convolution covers $n$ word images and outputs a scalar. By iterating this process continuously for each $n$-gram along with the sentence or article with multiple kernels, we obtain a 2-dimensional feature map. A subsequent 1-dimensional max-over-time pooling is applied to this feature map, and three fully-connected layers are used for conducting text classification finally. Experiments of several text classification datasets demonstrate surprisingly superior performances using the proposed model in comparison with existing methods.
We address the question of repeatedly learning linear classifiers against agents who are strategically trying to game the deployed classifiers, and we use the Stackelberg regret to measure the performance of our algorithms. First, we show that Stackelberg and external regret for the problem of strategic classification are strongly incompatible: i.e., there exist worst-case scenarios, where any sequence of actions providing sublinear external regret might result in linear Stackelberg regret and vice versa. Second, we present a strategy-aware algorithm for minimizing the Stackelberg regret for which we prove nearly matching upper and lower regret bounds. Finally, we provide simulations to complement our theoretical analysis. Our results advance the growing literature of learning from revealed preferences, which has so far focused on "smoother" assumptions from the perspective of the learner and the agents respectively.
Large pre-trained sentence encoders like BERT start a new chapter in natural language processing. A common practice to apply pre-trained BERT to sequence classification tasks (e.g., classification of sentences or sentence pairs) is by feeding the embedding of [CLS] token (in the last layer) to a task-specific classification layer, and then fine tune the model parameters of BERT and classifier jointly. In this paper, we conduct systematic analysis over several sequence classification datasets to examine the embedding values of [CLS] token before the fine tuning phase, and present the biased embedding distribution issue---i.e., embedding values of [CLS] concentrate on a few dimensions and are non-zero centered. Such biased embedding brings challenge to the optimization process during fine-tuning as gradients of [CLS] embedding may explode and result in degraded model performance. We further propose several simple yet effective normalization methods to modify the [CLS] embedding during the fine-tuning. Compared with the previous practice, neural classification model with the normalized embedding shows improvements on several text classification tasks, demonstrates the effectiveness of our method.
Multilingual pretrained language models (such as multilingual BERT) have achieved impressive results for cross-lingual transfer. However, due to the constant model capacity, multilingual pre-training usually lags behind the monolingual competitors. In this work, we present two approaches to improve zero-shot cross-lingual classification, by transferring the knowledge from monolingual pretrained models to multilingual ones. Experimental results on two cross-lingual classification benchmarks show that our methods outperform vanilla multilingual fine-tuning.
Graph Neural Networks (GNNs) are efficient approaches to process graph-structured data. Modelling long-distance node relations is essential for GNN training and applications. However, conventional GNNs suffer from bad performance in modelling long-distance node relations due to limited-layer information propagation. Existing studies focus on building deep GNN architectures, which face the over-smoothing issue and cannot model node relations in particularly long distance. To address this issue, we propose to model long-distance node relations by simply relying on shallow GNN architectures with two solutions: (1) Implicitly modelling by learning to predict node pair relations (2) Explicitly modelling by adding edges between nodes that potentially have the same label. To combine our two solutions, we propose a model-agnostic training framework named HighwayGraph, which overcomes the challenge of insufficient labeled nodes by sampling node pairs from the training set and adopting the self-training method. Extensive experimental results show that our HighwayGraph achieves consistent and significant improvements over four representative GNNs on three benchmark datasets.
Self-supervised pre-training of transformer models has shown enormous success in improving performance on a number of downstream tasks. However, fine-tuning on a new task still requires large amounts of task-specific labelled data to achieve good performance. We consider this problem of learning to generalize to new tasks with few examples as a meta-learning problem. While meta-learning has shown tremendous progress in recent years, its application is still limited to simulated problems or problems with limited diversity across tasks. We develop a novel method, LEOPARD, which enables optimization-based meta-learning across tasks with different number of classes, and evaluate different methods on generalization to diverse NLP classification tasks. LEOPARD is trained with the state-of-the-art transformer architecture and shows better generalization to tasks not seen at all during training, with as few as 4 examples per label. Across 17 NLP tasks, including diverse domains of entity typing, natural language inference, sentiment analysis, and several other text classification tasks, we show that LEOPARD learns better initial parameters for few-shot learning than self-supervised pre-training or multi-task training, outperforming many strong baselines, for example, yielding 14.5% average relative gain in accuracy on unseen tasks with only 4 examples per label.
Fake news has altered society in negative ways in politics and culture. It has adversely affected both online social network systems as well as offline communities and conversations. Using automatic machine learning classification models is an efficient way to combat the widespread dissemination of fake news. However, a lack of effective, comprehensive datasets has been a problem for fake news research and detection model development. Prior fake news datasets do not provide multimodal text and image data, metadata, comment data, and fine-grained fake news categorization at the scale and breadth of our dataset. We present Fakeddit, a novel multimodal dataset consisting of over 1 million samples from multiple categories of fake news. After being processed through several stages of review, the samples are labeled according to 2-way, 3-way, and 6-way classification categories through distant supervision. We construct hybrid text+image models and perform extensive experiments for multiple variations of classification, demonstrating the importance of the novel aspect of multimodality and fine-grained classification unique to Fakeddit.
We present a library to automatically embed signal processing and neural network predictions into the material robots are made of. Deep and shallow neural network models are first trained offline using state-of-the-art machine learning tools and then transferred onto general purpose microcontrollers that are co-located with a robot's sensors and actuators. We validate this approach using multiple examples: a smart robotic tire for terrain classification, a robotic finger sensor for load classification and a smart composite capable of regressing impact source localization. In each example, sensing and computation are embedded inside the material, creating artifacts that serve as stand-in replacement for otherwise inert conventional parts. The open source software library takes as inputs trained model files from higher level learning software, such as Tensorflow/Keras, and outputs code that is readable in a microcontroller that supports C. We compare the performance of this approach for various embedded platforms. In particular, we show that low-cost off-the-shelf microcontrollers can match the accuracy of a desktop computer, while being fast enough for real-time applications at different neural network configurations. We provide means to estimate the maximum number of parameters that the hardware will support based on the microcontroller's specifications.
Devices capable of detecting and categorizing acoustic scenes have numerous applications such as providing context-aware user experiences. In this paper, we address the task of characterizing acoustic scenes in a workplace setting from audio recordings collected with wearable microphones. The acoustic scenes, tracked with Bluetooth transceivers, vary dynamically with time from the egocentric perspective of a mobile user. Our dataset contains experience sampled long audio recordings collected from clinical providers in a hospital, who wore the audio badges during multiple work shifts. To handle the long egocentric recordings, we propose a Time Delay Neural Network~(TDNN)-based segment-level modeling. The experiments show that TDNN outperforms other models in the acoustic scene classification task. We investigate the effect of primary speaker's speech in determining acoustic scenes from audio badges, and provide a comparison between performance of different models. Moreover, we explore the relationship between the sequence of acoustic scenes experienced by the users and the nature of their jobs, and find that the scene sequence predicted by our model tend to possess similar relationship. The initial promising results reveal numerous research directions for acoustic scene classification via wearable devices as well as egocentric analysis of dynamic acoustic scenes encountered by the users.
We analyze the trade-off between model complexity and accuracy for random forests by breaking the trees up into individual classification rules and selecting a subset of them. We show experimentally that already a few rules are sufficient to achieve an acceptable accuracy close to that of the original model. Moreover, our results indicate that in many cases, this can lead to simpler models that clearly outperform the original ones.
Electroencephalography (EEG) data are difficult to obtain due to complex experimental setups and reduced comfort with prolonged wearing. This poses challenges to train powerful deep learning model with the limited EEG data. Being able to generate EEG data computationally could address this limitation. We propose a novel Wasserstein Generative Adversarial Network with gradient penalty (WGAN-GP) to synthesize EEG data. This network addresses several modeling challenges of simulating time-series EEG data including frequency artifacts and training instability. We further extended this network to a class-conditioned variant that also includes a classification branch to perform event-related classification. We trained the proposed networks to generate one and 64-channel data resembling EEG signals routinely seen in a rapid serial visual presentation (RSVP) experiment and demonstrated the validity of the generated samples. We also tested intra-subject cross-session classification performance for classifying the RSVP target events and showed that class-conditioned WGAN-GP can achieve improved event-classification performance over EEGNet.
Human movements are characterized by highly non-linear and multi-dimensional interactions within the motor system. Recently, an increasing emphasis on machine-learning applications has led to a significant contribution to the field of gait analysis, e.g., in increasing the classification performance. In order to ensure the generalizability of the machine-learning models, different data preprocessing steps are usually carried out to process the measured raw data before the classifications. In the past, various methods have been used for each of these preprocessing steps. However, there are hardly any standard procedures or rather systematic comparisons of these different methods and their impact on the classification performance. Therefore, the aim of this analysis is to compare different combinations of commonly applied data preprocessing steps and test their effects on the classification performance of gait patterns. A publicly available dataset on intra-individual changes of gait patterns was used for this analysis. Forty-two healthy participants performed 6 sessions of 15 gait trials for 1 day. For each trial, two force plates recorded the 3D ground reaction forces (GRFs). The data was preprocessed with the following steps: GRF filtering, time derivative, time normalization, data reduction, weight normalization and data scaling. Subsequently, combinations of all methods from each preprocessing step were analyzed by comparing their prediction performance in a six-session classification using Support Vector Machines, Random Forest Classifiers, Multi-Layer Perceptrons, and Convolutional Neural Networks. In conclusion, the present results provide first domain-specific recommendations for commonly applied data preprocessing methods and might help to build more comparable and more robust classification models based on machine learning that are suitable for a practical application.
We present Noisy Student Training, a semi-supervised learning approach that works well even when labeled data is abundant. Noisy Student Training achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. Noisy Student Training extends the idea of self-training and distillation with the use of equal-or-larger student models and noise added to the student during learning. On ImageNet, we first train an EfficientNet model on labeled images and use it as a teacher to generate pseudo labels for 300M unlabeled images. We then train a larger EfficientNet as a student model on the combination of labeled and pseudo labeled images. We iterate this process by putting back the student as the teacher. During the learning of the student, we inject noise such as dropout, stochastic depth, and data augmentation via RandAugment to the student so that the student generalizes better than the teacher. Models are available at https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet. Code is available at https://github.com/google-research/noisystudent.
Recent years have witnessed the emergence and increasing popularity of 3D medical imaging techniques with the development of 3D sensors and technology. However, achieving geometric invariance in the processing of 3D medical images is computationally expensive but nonetheless essential due to the presence of possible errors caused by rigid registration techniques. An alternative way to analyze medical imaging is by understanding the 3D shapes represented in terms of point-cloud. Though in the medical imaging community, 3D point-cloud processing is not a "go-to" choice, it is a canonical way to preserve rotation invariance. Unfortunately, due to the presence of discrete topology, one can not use the standard convolution operator on point-cloud. To the best of our knowledge, the existing ways to do "convolution" can not preserve the rotation invariance without explicit data augmentation. Therefore, we propose a rotation invariant convolution operator by inducing topology from hypersphere. Experimental validation has been performed on publicly available OASIS dataset in terms of classification accuracy between subjects with (without) dementia, demonstrating the usefulness of our proposed method in terms of model complexity, classification accuracy, and last but most important invariance to rotations.
Background and Objective: It is commonly accepted that accurate monitoring of neurodegenerative diseases is crucial for effective disease management and delivery of medication and treatment. This research develops automatic clinical monitoring techniques for PD, following treatment, using the novel application of EAs. Specifically, the research question addressed was: Can accurate monitoring of PD be achieved using EAs on rs-fMRI data for patients prescribed Modafinil (typically prescribed for PD patients to relieve physical fatigue)? Methods: This research develops novel clinical monitoring tools using data from a controlled experiment where participants were administered Modafinil versus placebo, examining the novel application of EAs to both map and predict the functional connectivity in participants using rs-fMRI data. Specifically, CGP was used to classify DCM analysis and timeseries data. Results were validated with two other commonly used classification methods (ANN and SVM) and via k-fold cross-validation. Results: Findings revealed a maximum accuracy of 74.57% for CGP. Furthermore, CGP provided comparable performance accuracy relative to ANN and SVM. Nevertheless, EAs enable us to decode the classifier, in terms of understanding the data inputs that are used, more easily than in ANN and SVM. Conclusions: These findings underscore the applicability of both DCM analyses for classification and CGP as a novel classification technique for brain imaging data with medical implications for medication monitoring. Furthermore, classification of fMRI data for research typically involves statistical modelling techniques being often hypothesis driven, whereas EAs use data-driven explanatory modelling methods resulting in numerous benefits. DCM analysis is novel for classification and advantageous as it provides information on the causal links between different brain regions.
By describing the features and abstractions of our world, language is a crucial tool for human learning and a promising source of supervision for machine learning models. We use language to improve few-shot visual classification in the underexplored scenario where natural language task descriptions are available during training, but unavailable for novel tasks at test time. Existing models for this setting sample new descriptions at test time and use those to classify images. Instead, we propose language-shaped learning (LSL), an end-to-end model that regularizes visual representations to predict language. LSL is conceptually simpler, more data efficient, and outperforms baselines in two challenging few-shot domains.
Machine learning models have made many decision support systems to be faster, more accurate, and more efficient. However, applications of machine learning in network security face a more disproportionate threat of active adversarial attacks compared to other domains. This is because machine learning applications in network security such as malware detection, intrusion detection, and spam filtering are by themselves adversarial in nature. In what could be considered an arm's race between attackers and defenders, adversaries constantly probe machine learning systems with inputs that are explicitly designed to bypass the system and induce a wrong prediction. In this survey, we first provide a taxonomy of machine learning techniques, tasks, and depth. We then introduce a classification of machine learning in network security applications. Next, we examine various adversarial attacks against machine learning in network security and introduce two classification approaches for adversarial attacks in network security. First, we classify adversarial attacks in network security based on a taxonomy of network security applications. Secondly, we categorize adversarial attacks in network security into a problem space vs feature space dimensional classification model. We then analyze the various defenses against adversarial attacks on machine learning-based network security applications. We conclude by introducing an adversarial risk grid map and evaluating several existing adversarial attacks against machine learning in network security using the risk grid map. We also identify where each attack classification resides within the adversarial risk grid map.