Abstract: Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time.

Abstract: Model distillation is an effective and widely used technique to transfer knowledge from a teacher to a student network The typical application is to transfer from a powerful large network or ensemble to a small network, in order to meet the low-memory or fast execution requirements In this paper, we present a deep mutual learning (DML) strategy Different from the one-way transfer between a static pre-defined teacher and a student in model distillation, with DML, an ensemble of students learn collaboratively and teach each other throughout the training process Our experiments show that a variety of network architectures benefit from mutual learning and achieve compelling results on both category and instance recognition tasks Surprisingly, it is revealed that no prior powerful teacher network is necessary - mutual learning of a collection of simple student networks works, and moreover outperforms distillation from a more powerful yet static teacher

Abstract: Recent work has shown that depth estimation from a stereo pair of images can be formulated as a supervised learning task to be resolved with convolutional neural networks (CNNs). However, current architectures rely on patch-based Siamese networks, lacking the means to exploit context information for finding correspondence in ill-posed regions. To tackle this problem, we propose PSMNet, a pyramid stereo matching network consisting of two main modules: spatial pyramid pooling and 3D CNN. The spatial pyramid pooling module takes advantage of the capacity of global context information by aggregating context in different scales and locations to form a cost volume. The 3D CNN learns to regularize cost volume using stacked multiple hourglass networks in conjunction with intermediate supervision. The proposed approach was evaluated on several benchmark datasets. Our method ranked first in the KITTI 2012 and 2015 leaderboards before March 18, 2018. The codes of PSMNet are available at: https://github.com/JiaRenChang/PSMNet.

Abstract: Supervised learning techniques construct predictive models by learning from a large number of training examples, where each training example has a label indicating its ground-truth output. Though current techniques have achieved great success, it is noteworthy that in many tasks it is difficult to get strong supervision information like fully ground-truth labels due to the high cost of the data-labeling process. Thus, it is desirable for machine-learning techniques to work with weak supervision. This article reviews some research progress of weakly supervised learning, focusing on three typical types of weak supervision: incomplete supervision, where only a subset of training data is given with labels; inexact supervision, where the training data are given with only coarse-grained labels; and inaccurate supervision, where the given labels are not always ground-truth.

Abstract: Image reconstruction is essential for imaging applications across the physical and life sciences, including optical and radar systems, magnetic resonance imaging, X-ray computed tomography, positron emission tomography, ultrasound imaging and radio astronomy. During image acquisition, the sensor encodes an intermediate representation of an object in the sensor domain, which is subsequently reconstructed into an image by an inversion of the encoding function. Image reconstruction is challenging because analytic knowledge of the exact inverse transform may not exist a priori, especially in the presence of sensor non-idealities and noise. Thus, the standard reconstruction approach involves approximating the inverse function with multiple ad hoc stages in a signal processing chain, the composition of which depends on the details of each acquisition strategy, and often requires expert parameter tuning to optimize reconstruction performance. Here we present a unified framework for image reconstruction-automated transform by manifold approximation (AUTOMAP)-which recasts image reconstruction as a data-driven supervised learning task that allows a mapping between the sensor and the image domain to emerge from an appropriate corpus of training data. We implement AUTOMAP with a deep neural network and exhibit its flexibility in learning reconstruction transforms for various magnetic resonance imaging acquisition strategies, using the same network architecture and hyperparameters. We further demonstrate that manifold learning during training results in sparse representations of domain transforms along low-dimensional data manifolds, and observe superior immunity to noise and a reduction in reconstruction artefacts compared with conventional handcrafted reconstruction methods. In addition to improving the reconstruction performance of existing acquisition methodologies, we anticipate that AUTOMAP and other learned reconstruction approaches will accelerate the development of new acquisition strategies across imaging modalities.

Abstract: Anomaly detection is a classical problem in computer vision, namely the determination of the normal from the abnormal when datasets are highly biased towards one class (normal) due to the insufficient sample size of the other class (abnormal). While this can be addressed as a supervised learning problem, a significantly more challenging problem is that of detecting the unknown/unseen anomaly case that takes us instead into the space of a one-class, semi-supervised learning paradigm. We introduce such a novel anomaly detection model, by using a conditional generative adversarial network that jointly learns the generation of high-dimensional image space and the inference of latent space. Employing encoder-decoder-encoder sub-networks in the generator network enables the model to map the input image to a lower dimension vector, which is then used to reconstruct the generated output image. The use of the additional encoder network maps this generated image to its latent representation. Minimizing the distance between these images and the latent vectors during training aids in learning the data distribution for the normal samples. As a result, a larger distance metric from this learned data distribution at inference time is indicative of an outlier from that distribution—an anomaly. Experimentation over several benchmark datasets, from varying domains, shows the model efficacy and superiority over previous state-of-the-art approaches.

Abstract: We present Deep Graph Infomax (DGI), a general approach for learning node representations within graph-structured data in an unsupervised manner. DGI relies on maximizing mutual information between patch representations and corresponding high-level summaries of graphs---both derived using established graph convolutional network architectures. The learnt patch representations summarize subgraphs centered around nodes of interest, and can thus be reused for downstream node-wise learning tasks. In contrast to most prior approaches to unsupervised learning with GCNs, DGI does not rely on random walk objectives, and is readily applicable to both transductive and inductive learning setups. We demonstrate competitive performance on a variety of node classification benchmarks, which at times even exceeds the performance of supervised learning.

Abstract: As a promising area in machine learning, multi-task learning (MTL) aims to improve the performance of multiple related learning tasks by leveraging useful information among them. In this paper, we give an overview of MTL by first giving a definition of MTL. Then several different settings of MTL are introduced, including multi-task supervised learning, multi-task unsupervised learning, multi-task semi-supervised learning, multi-task active learning, multi-task reinforcement learning, multi-task online learning and multi-task multi-view learning. For each setting, representative MTL models are presented. In order to speed up the learning process, parallel and distributed MTL models are introduced. Many areas, including computer vision, bioinformatics, health informatics, speech, natural language processing, web applications and ubiquitous computing, use MTL to improve the performance of the applications involved and some representative works are reviewed. Finally, recent theoretical analyses for MTL are presented.

Abstract: Speech separation is the task of separating target speech from background interference. Traditionally, speech separation is studied as a signal processing problem. A more recent approach formulates speech separation as a supervised learning problem, where the discriminative patterns of speech, speakers, and background noise are learned from training data. Over the past decade, many supervised separation algorithms have been put forward. In particular, the recent introduction of deep learning to supervised speech separation has dramatically accelerated progress and boosted separation performance. This paper provides a comprehensive overview of the research on deep learning based supervised speech separation in the last several years. We first introduce the background of speech separation and the formulation of supervised separation. Then, we discuss three main components of supervised separation: learning machines, training targets, and acoustic features. Much of the overview is on separation algorithms where we review monaural methods, including speech enhancement (speech-nonspeech separation), speaker separation (multitalker separation), and speech dereverberation, as well as multimicrophone techniques. The important issue of generalization, unique to supervised learning, is discussed. This overview provides a historical perspective on how advances are made. In addition, we discuss a number of conceptual issues, including what constitutes the target source.

Abstract: Deep neural networks have been shown to be very powerful modeling tools for many supervised learning tasks involving complex input patterns. However, they can also easily overfit to training set biases and label noises. In addition to various regularizers, example reweighting algorithms are popular solutions to these problems, but they require careful tuning of additional hyperparameters, such as example mining schedules and regularization hyperparameters. In contrast to past reweighting methods, which typically consist of functions of the cost value of each example, in this work we propose a novel meta-learning algorithm that learns to assign weights to training examples based on their gradient directions. To determine the example weights, our method performs a meta gradient descent step on the current mini-batch example weights (which are initialized from zero) to minimize the loss on a clean unbiased validation set. Our proposed method can be easily implemented on any type of deep network, does not require any additional hyperparameter tuning, and achieves impressive performance on class imbalance and corrupted label problems where only a small amount of clean validation data is available.

Abstract: Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on point clouds such as classification and segmentation. In this work, a novel end-to-end deep auto-encoder is proposed to address unsupervised learning challenges on point clouds. On the encoder side, a graph-based enhancement is enforced to promote local structures on top of PointNet. Then, a novel folding-based decoder deforms a canonical 2D grid onto the underlying 3D object surface of a point cloud, achieving low reconstruction errors even for objects with delicate structures. The proposed decoder only uses about 7% parameters of a decoder with fully-connected neural networks, yet leads to a more discriminative representation that achieves higher linear SVM classification accuracy than the benchmark. In addition, the proposed decoder structure is shown, in theory, to be a generic architecture that is able to reconstruct an arbitrary point cloud from a 2D grid. Our code is available at http://www.merl.com/research/license#FoldingNet

Abstract: Multiple instance learning (MIL) is a variation of supervised learning where a single class label is assigned to a bag of instances. In this paper, we state the MIL problem as learning the Bernoulli distribution of the bag label where the bag label probability is fully parameterized by neural networks. Furthermore, we propose a neural network-based permutation-invariant aggregation operator that corresponds to the attention mechanism. Notably, an application of the proposed attention-based operator provides insight into the contribution of each instance to the bag label. We show empirically that our approach achieves comparable performance to the best MIL methods on benchmark MIL datasets and it outperforms other methods on a MNIST-based MIL dataset and two real-life histopathology datasets without sacrificing interpretability.

Abstract: Deep neural networks excel at learning the training data, but often provide incorrect and confident predictions when evaluated on slightly different test examples. This includes distribution shifts, outliers, and adversarial examples. To address these issues, we propose Manifold Mixup, a simple regularizer that encourages neural networks to predict less confidently on interpolations of hidden representations. Manifold Mixup leverages semantic interpolations as additional training signal, obtaining neural networks with smoother decision boundaries at multiple levels of representation. As a result, neural networks trained with Manifold Mixup learn class-representations with fewer directions of variance. We prove theory on why this flattening happens under ideal conditions, validate it on practical situations, and connect it to previous works on information theory and generalization. In spite of incurring no significant computation and being implemented in a few lines of code, Manifold Mixup improves strong baselines in supervised learning, robustness to single-step adversarial attacks, and test log-likelihood.

Abstract: This paper studies the problem of learning image semantic segmentation networks only using image-level labels as supervision, which is important since it can significantly reduce human annotation efforts. Recent state-of-the-art methods on this problem first infer the sparse and discriminative regions for each object class using a deep classification network, then train semantic a segmentation network using the discriminative regions as supervision. Inspired by the traditional image segmentation methods of seeded region growing, we propose to train a semantic segmentation network starting from the discriminative regions and progressively increase the pixel-level supervision using by seeded region growing. The seeded region growing module is integrated in a deep segmentation network and can benefit from deep features. Different from conventional deep networks which have fixed/static labels, the proposed weakly-supervised network generates new labels using the contextual information within an image. The proposed method significantly outperforms the weakly-supervised semantic segmentation methods using static labels, and obtains the state-of-the-art performance, which are 63.2% mIoU score on the PASCAL VOC 2012 test set and 26.0% mIoU score on the COCO dataset.

Abstract: Most existing person re-identification (re-id) methods require supervised model learning from a separate large set of pairwise labelled training data for every single camera pair. This significantly limits their scalability and usability in real-world large scale deployments with the need for performing re-id across many camera views. To address this scalability problem, we develop a novel deep learning method for transferring the labelled information of an existing dataset to a new unseen (unlabelled) target domain for person re-id without any supervised learning in the target domain. Specifically, we introduce an Transferable Joint Attribute-Identity Deep Learning (TJ-AIDL) for simultaneously learning an attribute-semantic and identity-discriminative feature representation space transferrable to any new (unseen) target domain for re-id tasks without the need for collecting new labelled training data from the target domain (i.e. unsupervised learning in the target domain). Extensive comparative evaluations validate the superiority of this new TJ-AIDL model for unsupervised person re-id over a wide range of state-of-the-art methods on four challenging benchmarks including VIPeR, PRID, Market-1501, and DukeMTMC-ReID.

Abstract: In modern industries, machine health monitoring systems (MHMS) have been applied wildly with the goal of realizing predictive maintenance including failures tracking, downtime reduction, and assets preservation. In the era of big machinery data, data-driven MHMS have achieved remarkable results in the detection of faults after the occurrence of certain failures (diagnosis) and prediction of the future working conditions and the remaining useful life (prognosis). The numerical representation for raw sensory data is the key stone for various successful MHMS. Conventional methods are the labor-extensive as they usually depend on handcrafted features, which require expert knowledge. Inspired by the success of deep learning methods that redefine representation learning from raw data, we propose local feature-based gated recurrent unit (LFGRU) networks. It is a hybrid approach that combines handcrafted feature design with automatic feature learning for machine health monitoring. First, features from windows of input time series are extracted. Then, an enhanced bidirectional GRU network is designed and applied on the generated sequence of local features to learn the representation. A supervised learning layer is finally trained to predict machine condition. Experiments on three machine health monitoring tasks: tool wear prediction, gearbox fault diagnosis, and incipient bearing fault detection verify the effectiveness and generalization of the proposed LFGRU.

Abstract: While deeper convolutional networks are needed to achieve maximum accuracy in visual perception tasks, for many inputs shallower networks are sufficient. We exploit this observation by learning to skip convolutional layers on a per-input basis. We introduce SkipNet, a modified residual network, that uses a gating network to selectively skip convolutional blocks based on the activations of the previous layer. We formulate the dynamic skipping problem in the context of sequential decision making and propose a hybrid learning algorithm that combines supervised learning and reinforcement learning to address the challenges of non-differentiable skipping decisions. We show SkipNet reduces computation by \(30-90\%\) while preserving the accuracy of the original model on four benchmark datasets and outperforms the state-of-the-art dynamic networks and static compression methods. We also qualitatively evaluate the gating policy to reveal a relationship between image scale and saliency and the number of layers skipped.

Abstract: As robots and other intelligent agents move from simple environments and problems to more complex, unstructured settings, manually programming their behavior has become increasingly challenging and expensive. Often, it is easier for a teacher to demonstrate a desired behavior rather than attempt to manually engineer it. This process of learning from demonstrations, and the study of algorithms to do so, is called imitation learning. This work provides an introduction to imitation learning. It covers the underlying assumptions, approaches, and how they relate; the rich set of algorithms developed to tackle the problem; and advice on effective tools and implementation. We intend this paper to serve two audiences. First, we want to familiarize machine learning experts with the challenges of imitation learning, particularly those arising in robotics, and the interesting theoretical and practical distinctions between it and more familiar frameworks like statistical supervised learning theory and reinforcement learning. Second, we want to give roboticists and experts in applied artificial intelligence a broader appreciation for the frameworks and tools available for imitation learning. We pay particular attention to the intimate connection between imitation learning approaches and those of structured prediction Daume III et al. [2009]. To structure this discussion, we categorize imitation learning techniques based on the following key criteria which drive algorithmic decisions: 1) The structure of the policy space. Is the learned policy a time-index trajectory (trajectory learning), a mapping from observations to actions (so called behavioral cloning [Bain and Sammut, 1996]), or the result of a complex optimization or planning problem at each execution as is common in inverse optimal control methods [Kalman, 1964, Moylan and Anderson, 1973]. 2) The information available during training and testing. In particular, is the learning algorithm privy to the full state that the teacher possess? Is the learner able to interact with the teacher and gather corrections or more data? Does the learner have a (typically a priori) model of the system with which it interacts? Does the learner have access to the reward (cost) function that the teacher is attempting to optimize? 3) The notion of success. Different algorithmic approaches provide varying guarantees on the resulting learned behavior. These guarantees range from weaker (e.g., measuring disagreement with the agent’s decision) to stronger (e.g., providing guarantees on the performance of the learner with respect to a true cost function, either known or unknown). We organize our work by paying particular attention to distinction (1): dividing imitation learning into directly replicating desired behavior (sometimes called behavioral cloning) and learning the hidden objectives of the desired behavior from demonstrations (called inverse optimal control or inverse reinforcement learning [Russell, 1998]). In the latter case, behavior arises as the result of an optimization problem solved for each new instance that the learner faces. In addition to method analysis, we discuss the design decisions a practitioner must make when selecting an imitation learning approach. Moreover, application examples—such as robots that play table tennis [Kober and Peters, 2009], programs that play the game of Go [Silver et al., 2016], and systems that understand natural language [Wen et al., 2015]— illustrate the properties and motivations behind different forms of imitation learning. We conclude by presenting a set of open questions and point towards possible future research directions for machine learning.

Abstract: Model validation is the most important part of building a supervised model. For building a model with good generalization performance one must have a sensible data splitting strategy, and this is crucial for model validation. In this study, we conducted a comparative study on various reported data splitting methods. The MixSim model was employed to generate nine simulated datasets with different probabilities of mis-classification and variable sample sizes. Then partial least squares for discriminant analysis and support vector machines for classification were applied to these datasets. Data splitting methods tested included variants of cross-validation, bootstrapping, bootstrapped Latin partition, Kennard-Stone algorithm (K-S) and sample set partitioning based on joint X–Y distances algorithm (SPXY). These methods were employed to split the data into training and validation sets. The estimated generalization performances from the validation sets were then compared with the ones obtained from the blind test sets which were generated from the same distribution but were unseen by the training/validation procedure used in model construction. The results showed that the size of the data is the deciding factor for the qualities of the generalization performance estimated from the validation set. We found that there was a significant gap between the performance estimated from the validation set and the one from the test set for the all the data splitting methods employed on small datasets. Such disparity decreased when more samples were available for training/validation, and this is because the models were then moving towards approximations of the central limit theory for the simulated datasets used. We also found that having too many or too few samples in the training set had a negative effect on the estimated model performance, suggesting that it is necessary to have a good balance between the sizes of training set and validation set to have a reliable estimation of model performance. We also found that systematic sampling method such as K-S and SPXY generally had very poor estimation of the model performance, most likely due to the fact that they are designed to take the most representative samples first and thus left a rather poorly representative sample set for model performance estimation.

Abstract: Neural networks dominate the modern machine learning landscape, but their training and success still suffer from sensitivity to empirical choices of hyperparameters such as model architecture, loss function, and optimisation algorithm. In this work we present \emph{Population Based Training (PBT)}, a simple asynchronous optimisation algorithm which effectively utilises a fixed computational budget to jointly optimise a population of models and their hyperparameters to maximise performance. Importantly, PBT discovers a schedule of hyperparameter settings rather than following the generally sub-optimal strategy of trying to find a single fixed set to use for the whole course of training. With just a small modification to a typical distributed hyperparameter training framework, our method allows robust and reliable training of models. We demonstrate the effectiveness of PBT on deep reinforcement learning problems, showing faster wall-clock convergence and higher final performance of agents by optimising over a suite of hyperparameters. In addition, we show the same method can be applied to supervised learning for machine translation, where PBT is used to maximise the BLEU score directly, and also to training of Generative Adversarial Networks to maximise the Inception score of generated images. In all cases PBT results in the automatic discovery of hyperparameter schedules and model selection which results in stable training and better final performance.

Abstract: Weakly supervised learning with only coarse labels can obtain visual explanations of deep neural network such as attention maps by back-propagating gradients. These attention maps are then available as priors for tasks such as object localization and semantic segmentation. In one common framework we address three shortcomings of previous approaches in modeling such attention maps: We (1) make attention maps an explicit and natural component of the end-to-end training for the first time, (2) provide self-guidance directly on these maps by exploring supervision from the network itself to improve them, and (3) seamlessly bridge the gap between using weak and extra supervision if available. Despite its simplicity, experiments on the semantic segmentation task demonstrate the effectiveness of our methods. We clearly surpass the state-of-the-art on PASCAL VOC 2012 test and val. sets. Besides, the proposed framework provides a way not only explaining the focus of the learner but also feeding back with direct guidance towards specific tasks. Under mild assumptions our method can also be understood as a plug-in to existing weakly supervised learners to improve their generalization performance.

Abstract: We introduce a quantum neural network, QNN, that can represent labeled data, classical or quantum, and be trained by supervised learning. The quantum circuit consists of a sequence of parameter dependent unitary transformations which acts on an input quantum state. For binary classification a single Pauli operator is measured on a designated readout qubit. The measured output is the quantum neural network's predictor of the binary label of the input state. First we look at classifying classical data sets which consist of n-bit strings with binary labels. The input quantum state is an n-bit computational basis state corresponding to a sample string. We show how to design a circuit made from two qubit unitaries that can correctly represent the label of any Boolean function of n bits. For certain label functions the circuit is exponentially long. We introduce parameter dependent unitaries that can be adapted by supervised learning of labeled data. We study an example of real world data consisting of downsampled images of handwritten digits each of which has been labeled as one of two distinct digits. We show through classical simulation that parameters can be found that allow the QNN to learn to correctly distinguish the two data sets. We then discuss presenting the data as quantum superpositions of computational basis states corresponding to different label values. Here we show through simulation that learning is possible. We consider using our QNN to learn the label of a general quantum state. By example we show that this can be done. Our work is exploratory and relies on the classical simulation of small quantum systems. The QNN proposed here was designed with near-term quantum processors in mind. Therefore it will be possible to run this QNN on a near term gate model quantum computer where its power can be explored beyond what can be explored with simulation.

Abstract: A vast majority of computation in the brain is performed by spiking neural networks. Despite the ubiquity of such spiking, we currently lack an understanding of how biological spiking neural circuits learn and compute in vivo, as well as how we can instantiate such capabilities in artificial spiking circuits in silico. Here we revisit the problem of supervised learning in temporally coding multilayer spiking neural networks. First, by using a surrogate gradient approach, we derive SuperSpike, a nonlinear voltage-based three-factor learning rule capable of training multilayer networks of deterministic integrate-and-fire neurons to perform nonlinear computations on spatiotemporal spike patterns. Second, inspired by recent results on feedback alignment, we compare the performance of our learning rule under different credit assignment strategies for propagating output errors to hidden units. Specifically, we test uniform, symmetric, and random feedback, finding that simpler tasks can be solved with any type of feedback, while more complex tasks require symmetric feedback. In summary, our results open the door to obtaining a better scientific understanding of learning and computation in spiking neural networks by advancing our ability to train them to solve nonlinear problems involving transformations between different spatiotemporal spike time patterns.

Abstract: Multiple instance learning (MIL) is a variation of supervised learning where a single class label is assigned to a bag of instances. In this paper, we state the MIL problem as learning the Bernoulli distribution of the bag label where the bag label probability is fully parameterized by neural networks. Furthermore, we propose a neural network-based permutation-invariant aggregation operator that corresponds to the attention mechanism. Notably, an application of the proposed attention-based operator provides insight into the contribution of each instance to the bag label. We show empirically that our approach achieves comparable performance to the best MIL methods on benchmark MIL datasets and it outperforms other methods on a MNIST-based MIL dataset and two real-life histopathology datasets without sacrificing interpretability.

Abstract: Given the unprecedented availability of data and computing resources, there is widespread renewed interest in applying data-driven machine learning methods to problems for which the development of conventional engineering solutions is challenged by modeling or algorithmic deficiencies. This tutorial-style paper starts by addressing the questions of why and when such techniques can be useful. It then provides a high-level introduction to the basics of supervised and unsupervised learning. For both supervised and unsupervised learning, exemplifying applications to communication networks are discussed by distinguishing tasks carried out at the edge and at the cloud segments of the network at different layers of the protocol stack, with an emphasis on the physical layer.

Abstract: We consider the single image super-resolution problem in a more general case that the low-/high-resolution pairs and the down-sampling process are unavailable. Different from traditional super-resolution formulation, the low-resolution input is further degraded by noises and blurring. This complicated setting makes supervised learning and accurate kernel estimation impossible. To solve this problem, we resort to unsupervised learning without paired data, inspired by the recent successful image-to-image translation applications. With generative adversarial networks (GAN) as the basic component, we propose a Cycle-in-Cycle network structure to tackle the problem within three steps. First, the noisy and blurry input is mapped to a noise-free low-resolution space. Then the intermediate image is up-sampled with a pre-trained deep model. Finally, we fine-tune the two modules in an end-to-end manner to get the high-resolution output. Experiments on NTIRE2018 datasets demonstrate that the proposed unsupervised method achieves comparable results as the state-of-the-art supervised models.

Abstract: Unsupervised representation learning algorithms such as word2vec and ELMo improve the accuracy of many supervised NLP models, mainly because they can take advantage of large amounts of unlabeled text. However, the supervised models only learn from task-specific labeled data during the main training phase. We therefore propose Cross-View Training (CVT), a semi-supervised learning algorithm that improves the representations of a Bi-LSTM sentence encoder using a mix of labeled and unlabeled data. On labeled examples, standard supervised learning is used. On unlabeled examples, CVT teaches auxiliary prediction modules that see restricted views of the input (e.g., only part of a sentence) to match the predictions of the full model seeing the whole input. Since the auxiliary modules and the full model share intermediate representations, this in turn improves the full model. Moreover, we show that CVT is particularly effective when combined with multi-task learning. We evaluate CVT on five sequence tagging tasks, machine translation, and dependency parsing, achieving state-of-the-art results.