We investigate how the final parameters found by stochastic gradient descent are influenced by over-parameterization. We generate families of models by increasing the number of channels in a base network, and then perform a large hyper-parameter search to study how the test error depends on learning rate, batch size, and network width. We find that the optimal SGD hyper-parameters are determined by a “normalized noise scale,” which is a function of the batch size, learning rate, and initialization conditions. In the absence of batch normalization, the optimal normalized noise scale is directly proportional to width. Wider networks, with their higher optimal noise scale, also achieve higher test accuracy. These observations hold for MLPs, ConvNets, and ResNets, and for two different parameterization schemes (“Standard” and “NTK”). We observe a similar trend with batch normalization for ResNets. Surprisingly, since the largest stable learning rate is bounded, the largest batch size consistent with the optimal normalized noise scale decreases as the width increases.
http://arxiv.org/abs/1905.03776
Neural networks are proven to be remarkably successful for classification and diagnosis in medical applications. However, the ambiguity in the decision-making process and the interpretability of the learned features is a matter of concern. In this work, we propose a method for improving the feature interpretability of neural network classifiers. Initially, we propose a baseline convolutional neural network with state of the art performance in terms of accuracy and weakly supervised localization. Subsequently, the loss is modified to integrate robustness to adversarial examples into the training process. In this work, feature interpretability is quantified via evaluating the weakly supervised localization using the ground truth bounding boxes. Interpretability is also visually assessed using class activation maps and saliency maps. The method is applied to NIH ChestX-ray14, the largest publicly available chest x-rays dataset. We demonstrate that the adversarially robust optimization paradigm improves feature interpretability both quantitatively and visually.
http://arxiv.org/abs/1905.03767
Can we perform an end-to-end music source separation with a variable number of sources using a deep learning model? We present an extension of the Wave-U-Net model which allows end-to-end monaural source separation with a non-fixed number of sources. Furthermore, we propose multiplicative conditioning with instrument labels at the bottleneck of the Wave-U-Net and show its effect on the separation results. This approach leads to other types of conditioning such as audio-visual source separation and score-informed source separation.
http://arxiv.org/abs/1811.01850
Recent years have witnessed some exciting developments in the domain of generating images from scene-based text descriptions. These approaches have primarily focused on generating images from a static text description and are limited to generating images in a single pass. They are unable to generate an image interactively based on an incrementally additive text description (something that is more intuitive and similar to the way we describe an image). We propose a method to generate an image incrementally based on a sequence of graphs of scene descriptions (scene-graphs). We propose a recurrent network architecture that preserves the image content generated in previous steps and modifies the cumulative image as per the newly provided scene information. Our model utilizes Graph Convolutional Networks (GCN) to cater to variable-sized scene graphs along with Generative Adversarial image translation networks to generate realistic multi-object images without needing any intermediate supervision during training. We experiment with Coco-Stuff dataset which has multi-object images along with annotations describing the visual scene and show that our model significantly outperforms other approaches on the same dataset in generating visually consistent images for incrementally growing scene graphs.
http://arxiv.org/abs/1905.03743
The interaction between the vestibular and ocular system has primarily been studied in controlled environments. Consequently, off-the shelf tools for categorization of gaze events (e.g. fixations, pursuits, saccade) fail when head movements are allowed. Our approach was to collect a novel, naturalistic, and multimodal dataset of eye+head movements when subjects performed everyday tasks while wearing a mobile eye tracker equipped with an inertial measurement unit and a 3D stereo camera. This Gaze-in-the-Wild dataset (GW) includes eye+head rotational velocities (deg/s), infrared eye images and scene imagery (RGB+D). A portion was labelled by coders into gaze motion events with a mutual agreement of 0.72 sample based Cohen’s $\kappa$. This labelled data was used to train and evaluate two machine learning algorithms, Random Forest and a Recurrent Neural Network model, for gaze event classification. Assessment involved the application of established and novel event based performance metrics. Classifiers achieve $\sim$90$\%$ human performance in detecting fixations and saccades but fall short (60$\%$) on detecting pursuit movements. Moreover, pursuit classification is far worse in the absence of head movement information. A subsequent analysis of feature significance in our best-performing model revealed a reliance upon absolute eye and head velocity, indicating that classification does not require spatial alignment of the head and eye tracking coordinate systems. The GW dataset, trained classifiers and evaluation metrics will be made publicly available with the intention of facilitating growth in the emerging area of head-free gaze event classification.
http://arxiv.org/abs/1905.13146
We study how Reinforcement Learning can be employed to optimally control parameters in evolutionary algorithms. We control the mutation probability of a (1+1) evolutionary algorithm on the OneMax function. This problem is modeled as a Markov Decision Process and solved with Value Iteration via the known transition probabilities. It is then solved via Q-Learning, a Reinforcement Learning algorithm, where the exact transition probabilities are not needed. This approach also allows previous expert or empirical knowledge to be included into learning. It opens new perspectives, both formally and computationally, for the problem of parameter control in optimization.
http://arxiv.org/abs/1905.03726
The Lambek calculus can be considered as a version of non-commutative intuitionistic linear logic. One of the interesting features of the Lambek calculus is the so-called “Lambek’s restriction,” that is, the antecedent of any provable sequent should be non-empty. In this paper we discuss ways of extending the Lambek calculus with the linear logic exponential modality while keeping Lambek’s restriction. Interestingly enough, we show that for any system equipped with a reasonable exponential modality the following holds: if the system enjoys cut elimination and substitution to the full extent, then the system necessarily violates Lambek’s restriction. Nevertheless, we show that two of the three conditions can be implemented. Namely, we design a system with Lambek’s restriction and cut elimination and another system with Lambek’s restriction and substitution. For both calculi we prove that they are undecidable, even if we take only one of the two divisions provided by the Lambek calculus. The system with cut elimination and substitution and without Lambek’s restriction is folklore and known to be undecidable.
http://arxiv.org/abs/1608.02254
This paper presents a data-efficient approach to learning transferable forward models for robotic push manipulation. Our approach extends our previous work on contact-based predictors by leveraging information on the pushed object’s local surface features. We test the hypothesis that, by conditioning predictions on local surface features, we can achieve generalisation across objects of different shapes. In doing so, we do not require a CAD model of the object but rather rely on a point cloud object model (PCOM). Our approach involves learning motion models that are specific to contact models. Contact models encode the contacts seen during training time and allow generating similar contacts at prediction time. Predicting on familiar ground reduces the motion models’ sample complexity while using local contact information for prediction increases their transferability. In extensive experiments in simulation, our approach is capable of transfer learning for various test objects, outperforming a baseline predictor. We support those results with a proof of concept on a real robot.
http://arxiv.org/abs/1905.03720
We present SOSELETO (SOurce SELEction for Target Optimization), a new method for exploiting a source dataset to solve a classification problem on a target dataset. SOSELETO is based on the following simple intuition: some source examples are more informative than others for the target problem. To capture this intuition, source samples are each given weights; these weights are solved for jointly with the source and target classification problems via a bilevel optimization scheme. The target therefore gets to choose the source samples which are most informative for its own classification task. Furthermore, the bilevel nature of the optimization acts as a kind of regularization on the target, mitigating overfitting. SOSELETO may be applied to both classic transfer learning, as well as the problem of training on datasets with noisy labels; we show state of the art results on both of these problems.
http://arxiv.org/abs/1805.09622
Reliable anticipation of pedestrian trajectory is imperative for the operation of autonomous vehicles and can significantly enhance the functionality of advanced driver assistance systems. While significant progress has been made in the field of pedestrian detection, forecasting pedestrian trajectories remains a challenging problem due to the unpredictable nature of pedestrians and the huge space of potentially useful features. In this work, we present a deep learning approach for pedestrian trajectory forecasting using a single vehicle-mounted camera. Deep learning models that have revolutionized other areas in computer vision have seen limited application to trajectory forecasting, in part due to the lack of richly annotated training data. We address the lack of training data by introducing a scalable machine annotation scheme that enables our model to be trained using a large dataset without human annotation. In addition, we propose Dynamic Trajectory Predictor (DTP), a model for forecasting pedestrian trajectory up to one second into the future. DTP is trained using both human and machine-annotated data, and anticipates dynamic motion that is not captured by linear models. Experimental evaluation confirms the benefits of the proposed model.
http://arxiv.org/abs/1905.03681
Convolutional networks for single-view object reconstruction have shown impressive performance and have become a popular subject of research. All existing techniques are united by the idea of having an encoder-decoder network that performs non-trivial reasoning about the 3D structure of the output space. In this work, we set up two alternative approaches that perform image classification and retrieval respectively. These simple baselines yield better results than state-of-the-art methods, both qualitatively and quantitatively. We show that encoder-decoder methods are statistically indistinguishable from these baselines, thus indicating that the current state of the art in single-view object reconstruction does not actually perform reconstruction but image classification. We identify aspects of popular experimental procedures that elicit this behavior and discuss ways to improve the current state of research.
http://arxiv.org/abs/1905.03678
The performance of deep neural networks improves with more annotated data. The problem is that the budget for annotation is limited. One solution to this is active learning, where a model asks human to annotate data that it perceived as uncertain. A variety of recent methods have been proposed to apply active learning to deep networks but most of them are either designed specific for their target tasks or computationally inefficient for large networks. In this paper, we propose a novel active learning method that is simple but task-agnostic, and works efficiently with the deep networks. We attach a small parametric module, named “loss prediction module,” to a target network, and learn it to predict target losses of unlabeled inputs. Then, this module can suggest data that the target model is likely to produce a wrong prediction. This method is task-agnostic as networks are learned from a single loss regardless of target tasks. We rigorously validate our method through image classification, object detection, and human pose estimation, with the recent network architectures. The results demonstrate that our method consistently outperforms the previous methods over the tasks.
http://arxiv.org/abs/1905.03677
In this paper, we are interested in boosting the representation capability of convolution neural networks which utilizing the inverted residual structure. Based on the success of Inverted Residual structure[Sandler et al. 2018] and Interleaved Low-Rank Group Convolutions[Sun et al. 2018], we rethink this two pattern of neural network structure, rather than NAS(Neural architecture search) method[Zoph and Le 2017; Pham et al. 2018; Liu et al. 2018b], we introduce uneven point-wise group convolution, which provide a novel search space for designing basic blocks to obtain better trade-off between representation capability and computational cost. Meanwhile, we propose two novel information flow patterns that will enable cross-group information flow for multiple group convolution layers with and without any channel permute/shuffle operation. Dense experiments on image classification task show that our proposed model, named Seesaw-Net, achieves state-of-the-art(SOTA) performance with limited computation and memory cost. Our code will be open-source and available together with pre-trained models.
http://arxiv.org/abs/1905.03672
This work tackles the problem of semi-supervised learning of image classifiers. Our main insight is that the field of semi-supervised learning can benefit from the quickly advancing field of self-supervised visual representation learning. Unifying these two approaches, we propose the framework of self-supervised semi-supervised learning ($S^4L$) and use it to derive two novel semi-supervised image classification methods. We demonstrate the effectiveness of these methods in comparison to both carefully tuned baselines, and existing semi-supervised learning methods. We then show that $S^4L$ and existing semi-supervised methods can be jointly trained, yielding a new state-of-the-art result on semi-supervised ILSVRC-2012 with 10% of labels.
http://arxiv.org/abs/1905.03670
Modern neural network training relies on piece-wise (sub-)differentiable functions in order to use backpropation for efficient calculation of gradients. In this work, we introduce a novel method to allow for non-differentiable functions at intermediary layers of deep neural networks. We do so through the introduction of a differentiable approximation bridge (DAB) neural network which provides smooth approximations to the gradient of the non-differentiable function. We present strong empirical results (performing over 600 experiments) in three different domains: unsupervised (image) representation learning, image classification, and sequence sorting to demonstrate that our proposed method improves state of the art performance. We demonstrate that utilizing non-differentiable functions in unsupervised (image) representation learning improves reconstruction quality and posterior linear separability by 10%. We also observe an accuracy improvement of 77% in neural sequence sorting and a 25% improvement against the straight-through estimator [3] in an image classification setting with the sort non-linearity. This work enables the usage of functions that were previously not usable in neural networks.
http://arxiv.org/abs/1905.03658
Stochastic optimization algorithms update models with cheap per-iteration costs sequentially, which makes them amenable for large-scale data analysis. Such algorithms have been widely studied for structured sparse models where the sparsity information is very specific, e.g., convex sparsity-inducing norms or $\ell^0$-norm. However, these norms cannot be directly applied to the problem of complex (non-convex) graph-structured sparsity models, which have important application in disease outbreak and social networks, etc. In this paper, we propose a stochastic gradient-based method for solving graph-structured sparsity constraint problems, not restricted to the least square loss. We prove that our algorithm enjoys a linear convergence up to a constant error, which is competitive with the counterparts in the batch learning setting. We conduct extensive experiments to show the efficiency and effectiveness of the proposed algorithms.
http://arxiv.org/abs/1905.03652
This work presents a text mining context and its use for a deep analysis of the messages delivered by the politicians. Specifically, we deal with an expert systems-based exploration of the rhetoric dynamics of a large collection of US Presidents’ speeches, ranging from Washington to Trump. In particular, speeches are viewed as complex expert systems whose structures can be effectively analyzed through rank-size laws. The methodological contribution of the paper is twofold. First, we develop a text mining-based procedure for the construction of the dataset by using a web scraping routine on the Miller Center website – the repository collecting the speeches. Second, we explore the implicit structure of the discourse data by implementing a rank-size procedure over the individual speeches, being the words of each speech ranked in terms of their frequencies. The scientific significance of the proposed combination of text-mining and rank-size approaches can be found in its flexibility and generality, which let it be reproducible to a wide set of expert systems and text mining contexts. The usefulness of the proposed method and the speech subsequent analysis is demonstrated by the findings themselves. Indeed, in terms of impact, it is worth noting that interesting conclusions of social, political and linguistic nature on how 45 United States Presidents, from April 30, 1789 till February 28, 2017 delivered political messages can be carried out. Indeed, the proposed analysis shows some remarkable regularities, not only inside a given speech, but also among different speeches. Moreover, under a purely methodological perspective, the presented contribution suggests possible ways of generating a linguistic decision-making algorithm.
http://arxiv.org/abs/1905.04705
At present, lesion segmentation is still performed manually (or semi-automatically) by medical experts. To facilitate this process, we contribute a fully-automatic lesion segmentation pipeline. This work proposes a method as a part of the LiTS (Liver Tumor Segmentation Challenge) competition for ISBI 17 and MICCAI 17 comparing methods for automatics egmentation of liver lesions in CT scans. By utilizing cascaded, densely connected 2D U-Nets and a Tversky-coefficient based loss function, our framework achieves very good shape extractions with high detection sensitivity, with competitive scores at time of publication. In addition, adjusting hyperparameters in our Tversky-loss allows to tune the network towards higher sensitivity or robustness.
http://arxiv.org/abs/1905.03639
We present a novel real-time, collaborative, and interactive AI painting system, Mappa Mundi, for artistic Mind Map creation. The system consists of a voice-based input interface, an automatic topic expansion module, and an image projection module. The key innovation is to inject Artificial Imagination into painting creation by considering lexical and phonological similarities of language, learning and inheriting artist’s original painting style, and applying the principles of Dadaism and impossibility of improvisation. Our system indicates that AI and artist can collaborate seamlessly to create imaginative artistic painting and Mappa Mundi has been applied in art exhibition in UCCA, Beijing
http://arxiv.org/abs/1905.03638
The following article introduces a new parametric synthesis algorithm for sound textures inspired by existing methods used for visual textures. Using a 2D Convolutional Neural Network (CNN), a sound signal is modified until the temporal cross-correlations of the feature maps of its log-spectrogram resemble those of a target texture. We show that the resulting synthesized sound signal is both different from the original and of high quality, while being able to reproduce singular events appearing in the original. This process is performed in the time domain, discarding the harmful phase recovery step which usually concludes synthesis performed in the time-frequency domain. It is also straightforward and flexible, as it does not require any fine tuning between several losses when synthesizing diverse sound textures. A way of extending the synthesis in order to produce a sound of any length is also presented, after which synthesized spectrograms and sound signals are showcased. We also discuss on the choice of CNN, on border effects in our synthesized signals and on possible ways of modifying the algorithm in order to improve its current long computation time.
http://arxiv.org/abs/1905.03637
Objects moving at high speed along complex trajectories often appear in videos, especially videos of sports. Such objects elapse non-negligible distance during exposure time of a single frame and therefore their position in the frame is not well defined. They appear as semi-transparent streaks due to the motion blur and cannot be reliably tracked by standard trackers. We propose a novel approach called Tracking by Deblatting based on the observation that motion blur is directly related to the intra-frame trajectory of an object. Blur is estimated by solving two intertwined inverse problems, blind deblurring and image matting, which we call deblatting. The trajectory is then estimated by fitting a piecewise quadratic curve, which models physically justifiable trajectories. As a result, tracked objects are precisely localized with higher temporal resolution than by conventional trackers. The proposed TbD tracker was evaluated on a newly created dataset of videos with ground truth obtained by a high-speed camera using a novel Trajectory-IoU metric that generalizes the traditional Intersection over Union and measures the accuracy of the intra-frame trajectory. The proposed method outperforms baseline both in recall and trajectory accuracy.
http://arxiv.org/abs/1905.03633
This paper addresses the problem of block-online processing for multi-channel speech enhancement. We consider several variants of a system that performs beamforming supported by DNN-based Voice Activity Detection followed by post-filtering. The speaker is targeted through estimating relative transfer functions between microphones. Each block of the input signals is processed independently in order to make the method applicable in highly dynamic environments. The performance loss caused by the short length of the processing block is studied and compared with results achieved when recordings are processed as one block (batch processing). The experimental evaluation of the proposed method is performed on large datasets of CHiME-4 and on another dataset featuring moving target speaker. The experiments are evaluated in terms of objective criteria and Word Error Rate achieved by a baseline Automatic Speech Recognition system, for which the enhancement method serves as a front-end solution. The results indicate that the proposed method is robust with respect to the length of the processing block and yields significant WER improvement even for a short block length of 250 ms.
http://arxiv.org/abs/1905.03632
In this paper, we are interested in exploiting textual and acoustic data of an utterance for the speech emotion classification task. The baseline approach models the information from audio and text independently using two deep neural networks (DNNs). The outputs from both the DNNs are then fused for classification. As opposed to using knowledge from both the modalities separately, we propose a framework to exploit acoustic information in tandem with lexical data. The proposed framework uses two bi-directional long short-term memory (BLSTM) for obtaining hidden representations of the utterance. Furthermore, we propose an attention mechanism, referred to as the multi-hop, which is trained to automatically infer the correlation between the modalities. The multi-hop attention first computes the relevant segments of the textual data corresponding to the audio signal. The relevant textual data is then applied to attend parts of the audio signal. To evaluate the performance of the proposed system, experiments are performed in the IEMOCAP dataset. Experimental results show that the proposed technique outperforms the state-of-the-art system by 6.5% relative improvement in terms of weighted accuracy.
http://arxiv.org/abs/1904.10788
The present study aims to investigate similarities between how humans and connectionist models experience difficulty in arithmetic problems. Problem difficulty was operationalized by the number of carries involved in solving a given problem. Problem difficulty was measured in humans by response time, and in models by computational steps. The present study found that both humans and connectionist models experience difficulty similarly when solving binary addition and subtraction. Specifically, both agents found difficulty to be strictly increasing with respect to the number of carries. Another notable similarity is that problem difficulty increases more steeply in subtraction than in addition, for both humans and connectionist models. Further investigation on two model hyperparameters — confidence threshold and hidden dimension — shows higher confidence thresholds cause the model to take more computational steps to arrive at the correct answer. Likewise, larger hidden dimensions cause the model to take more computational steps to correctly answer arithmetic problems; however, this effect by hidden dimensions is negligible.
http://arxiv.org/abs/1905.03617
Neural state-of-the-art sequence-to-sequence (seq2seq) models often do not perform well for small training sets. We address paradigm completion, the morphological task of, given a partial paradigm, generating all missing forms. We propose two new methods for the minimal-resource setting: (i) Paradigm transduction: Since we assume only few paradigms available for training, neural seq2seq models are able to capture relationships between paradigm cells, but are tied to the idiosyncracies of the training set. Paradigm transduction mitigates this problem by exploiting the input subset of inflected forms at test time. (ii) Source selection with high precision (SHIP): Multi-source models which learn to automatically select one or multiple sources to predict a target inflection do not perform well in the minimal-resource setting. SHIP is an alternative to identify a reliable source if training data is limited. On a 52-language benchmark dataset, we outperform the previous state of the art by up to 9.71% absolute accuracy.
http://arxiv.org/abs/1809.08733
We propose a real-time image fusion method using pre-trained neural networks. Our method generates a single image containing features from multiple sources. We first decompose images into a base layer representing large scale intensity variations, and a detail layer containing small scale changes. We use visual saliency to fuse the base layers, and deep feature maps extracted from a pre-trained neural network to fuse the detail layers. We conduct ablation studies to analyze our method’s parameters such as decomposition filters, weight construction methods, and network depth and architecture. Then, we validate its effectiveness and speed on thermal, medical, and multi-focus fusion. We also apply it to multiple image inputs such as multi-exposure sequences. The experimental results demonstrate that our technique achieves state-of-the-art performance in visual quality, objective assessment, and runtime efficiency.
http://arxiv.org/abs/1905.03590
Rewards are sparse in the real world and most of today’s reinforcement learning algorithms struggle with such sparsity. One solution to this problem is to allow the agent to create rewards for itself - thus making rewards dense and more suitable for learning. In particular, inspired by curious behaviour in animals, observing something novel could be rewarded with a bonus. Such bonus is summed up with the real task reward - making it possible for RL algorithms to learn from the combined reward. We propose a new curiosity method which uses episodic memory to form the novelty bonus. To determine the bonus, the current observation is compared with the observations in memory. Crucially, the comparison is done based on how many environment steps it takes to reach the current observation from those in memory - which incorporates rich information about environment dynamics. This allows us to overcome the known “couch-potato” issues of prior work - when the agent finds a way to instantly gratify itself by exploiting actions which lead to hardly predictable consequences. We test our approach in visually rich 3D environments in ViZDoom, DMLab and MuJoCo. In navigational tasks from ViZDoom and DMLab, our agent outperforms the state-of-the-art curiosity method ICM. In MuJoCo, an ant equipped with our curiosity module learns locomotion out of the first-person-view curiosity only.
http://arxiv.org/abs/1810.02274
Justification theory is a unifying framework for semantics of non-monotonic logics. It is built on the notion of a justification, which intuitively is a graph that explains the truth value of certain facts in a structure. Knowledge representation languages covered by justification theory include logic programs, argumentation frameworks, inductive definitions, and nested inductive and coinductive definitions. In addition, justifications are also used for implementation purposes. They are used to compute unfounded sets in modern ASP solvers, can be used to check for relevance of atoms in complete search algorithms, and recent lazy grounding algorithms are built on top of them. In this extended abstract, we lay out possible extensions to justification theory.
http://arxiv.org/abs/1905.06184
Foreseeing the future is one of the key factors of intelligence. It involves understanding of the past and current environment as well as decent experience of its possible dynamics. In this work, we address future prediction at the abstract level of activities. We propose a network module for learning embeddings of the environment’s dynamics in a self-supervised way. To take the ambiguities and high variances in the future activities into account, we use a multi-hypotheses scheme that can represent multiple futures. We demonstrate the approach by classifying future activities on the Epic-Kitchens and Breakfast datasets. Moreover, we generate captions that describe the future activities
http://arxiv.org/abs/1905.03578
In recent years, deep learning has presented a great advance in hyperspectral image (HSI) classification. Particularly, Long Short-Term Memory (LSTM), as a special deep learning structure, has shown great ability in modeling long-term dependencies in the time dimension of video or the spectral dimension of HSIs. However, the loss of spatial information makes it quite difficult to obtain the better performance. In order to address this problem, two novel deep models are proposed to extract more discriminative spatial-spectral features by exploiting the Convolutional LSTM (ConvLSTM) for the first time. By taking the data patch in a local sliding window as the input of each memory cell band by band, the 2-D extended architecture of LSTM is considered for building the spatial-spectral ConvLSTM 2-D Neural Network (SSCL2DNN) to model long-range dependencies in the spectral domain. To take advantage of spatial and spectral information more effectively for extracting a more discriminative spatial-spectral feature representation, the spatial-spectral ConvLSTM 3-D Neural Network (SSCL3DNN) is further proposed by extending LSTM to 3-D version. The experiments, conducted on three commonly used HSI data sets, demonstrate that the proposed deep models have certain competitive advantages and can provide better classification performance than other state-of-the-art approaches.
http://arxiv.org/abs/1905.03577
In this work we address the problem of finding reliable pixel-level correspondences under difficult imaging conditions. We propose an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector. By postponing the detection to a later stage, the obtained keypoints are more stable than their traditional counterparts based on early detection of low-level structures. We show that this model can be trained using pixel correspondences extracted from readily available large-scale SfM reconstructions, without any further annotations. The proposed method obtains state-of-the-art performance on both the difficult Aachen Day-Night localization dataset and the InLoc indoor localization benchmark, as well as competitive performance on other benchmarks for image matching and 3D reconstruction.
http://arxiv.org/abs/1905.03561
Supervised deep learning techniques have achieved great success in various fields due to getting rid of the limitation of handcrafted representations. However, most previous image retargeting algorithms still employ fixed design principles such as using gradient map or handcrafted features to compute saliency map, which inevitably restricts its generality. Deep learning techniques may help to address this issue, but the challenging problem is that we need to build a large-scale image retargeting dataset for the training of deep retargeting models. However, building such a dataset requires huge human efforts. In this paper, we propose a novel deep cyclic image retargeting approach, called Cycle-IR, to firstly implement image retargeting with a single deep model, without relying on any explicit user annotations. Our idea is built on the reverse mapping from the retargeted images to the given input images. If the retargeted image has serious distortion or excessive loss of important visual information, the reverse mapping is unlikely to restore the input image well. We constrain this forward-reverse consistency by introducing a cyclic perception coherence loss. In addition, we propose a simple yet effective image retargeting network (IRNet) to implement the image retargeting process. Our IRNet contains a spatial and channel attention layer, which is able to discriminate visually important regions of input images effectively, especially in cluttered images. Given arbitrary sizes of input images and desired aspect ratios, our Cycle-IR can produce visually pleasing target images directly. Extensive experiments on the standard RetargetMe dataset show the superiority of our Cycle-IR. In addition, our Cycle-IR outperforms the Multiop method and obtains the best result in the user study. Code is available at https://github.com/mintanwei/Cycle-IR.
http://arxiv.org/abs/1905.03556
During the last decade, Convolutional Neural Networks (CNNs) have become the de facto standard for various Computer Vision and Machine Learning operations. CNNs are feed-forward Artificial Neural Networks (ANNs) with alternating convolutional and subsampling layers. Deep 2D CNNs with many hidden layers and millions of parameters have the ability to learn complex objects and patterns providing that they can be trained on a massive size visual database with ground-truth labels. With a proper training, this unique ability makes them the primary tool for various engineering applications for 2D signals such as images and video frames. Yet, this may not be a viable option in numerous applications over 1D signals especially when the training data is scarce or application-specific. To address this issue, 1D CNNs have recently been proposed and immediately achieved the state-of-the-art performance levels in several applications such as personalized biomedical data classification and early diagnosis, structural health monitoring, anomaly detection and identification in power electronics and motor-fault detection. Another major advantage is that a real-time and low-cost hardware implementation is feasible due to the simple and compact configuration of 1D CNNs that perform only 1D convolutions (scalar multiplications and additions). This paper presents a comprehensive review of the general architecture and principals of 1D CNNs along with their major engineering applications, especially focused on the recent progress in this field. Their state-of-the-art performance is highlighted concluding with their unique properties. The benchmark datasets and the principal 1D CNN software used in those applications are also publically shared in a dedicated website.
http://arxiv.org/abs/1905.03554
In this paper, we propose a novel adaptive kernel for the radial basis function (RBF) neural networks. The proposed kernel adaptively fuses the Euclidean and cosine distance measures to exploit the reciprocating properties of the two. The proposed framework dynamically adapts the weights of the participating kernels using the gradient descent method thereby alleviating the need for predetermined weights. The proposed method is shown to outperform the manual fusion of the kernels on three major problems of estimation namely nonlinear system identification, pattern classification and function approximation.
http://arxiv.org/abs/1905.03546
Human-in-the-loop (HITL), which introduces human knowledge to machine learning, has been used in fine-grained recognition to estimate categories from the difference of local features. The conventional HITL approach has been successfully applied in non-deep machine learning, but it is difficult to use it with deep learning due to the enormous number of model parameters. To tackle this problem, in this paper, we propose using the Attention Branch Network (ABN) which is a visual explanation model. ABN applies an attention map for visual explanation to an attention mechanism. First, we manually modify the attention map obtained from ABN on the basis of human knowledge. Then, we use the modified attention map to an attention mechanism that enables ABN to adjust the recognition score. Second, for applying HITL to deep learning, we propose a fine-tuning approach that uses the modified attention map. Our fine-tuning updates the attention and perception branches of the ABN by using the training loss calculated from the attention map output from the ABN along with the modified attention map. This fine-tuning enables the ABN to output an attention map corresponding to human knowledge. Additionally, we use the updated attention map with its embedded human knowledge as an attention mechanism and inference at the perception branch, which improves the performance of ABN. Experimental results with the ImageNet dataset, CUB-200-2010 dataset, and IDRiD demonstrate that our approach clarifies the attention map in terms of visual explanation and improves the classification performance.
http://arxiv.org/abs/1905.03540
Currently, the text document retrieval systems have many challenges in exploring the semantics of queries and documents. Each query implies information which does not appear in the query but the documents related with the information are also expected by user. The disadvantage of the previous spreading activation algorithms could be many irrelevant concepts added to the query. In this paper, a proposed novel algorithm is only activate and add to the query named entities which are related with original entities in the query and explicit relations in the query.
http://arxiv.org/abs/1905.06114
As skin cancer is one of the most frequent cancers globally, accurate, non-invasive dermoscopy-based diagnosis becomes essential and promising. A task of the Part 3 of the ISIC Skin Image Analysis Challenge at MICCAI 2018 is to predict seven disease classes with skin lesion images, including melanoma (MEL), melanocytic nevus (NV), basal cell carcinoma (BCC), actinic keratosis / Bowen’s disease (intraepithelial carcinoma) (AKIEC), benign keratosis (solar lentigo / seborrheic keratosis / lichen planus-like keratosis) (BKL), dermatofibroma (DF) and vascular lesion (VASC) as defined by the International Dermatology Society. In this work, we design the WonDerM pipeline, that resamples the preprocessed skin lesion images, builds neural network architecture fine-tuned with segmentation task data (the Part 1), and uses an ensemble method to classify the seven skin diseases. Our model achieved the score 0.899, 0.785 in validation data and test data, respectively.
http://arxiv.org/abs/1808.03426
We consider bounded width CNF-formulas where the width is measured by popular graph width measures on graphs associated to CNF-formulas. Such restricted graph classes, in particular those of bounded treewidth, have been extensively studied for their uses in the design of algorithms for various computational problems on CNF-formulas. Here we consider the expressivity of these formulas in the model of clausal encodings with auxiliary variables. We first show that bounding the width for many of the measures from the literature leads to a dramatic loss of expressivity, restricting the formulas to such of low communication complexity. We then show that the width of optimal encodings with respect to different measures is strongly linked: there are two classes of width measures, one containing primal treewidth and the other incidence cliquewidth, such that in each class the width of optimal encodings only differs by constant factors. Moreover, between the two classes the width differs at most by a factor logarithmic in the number of variables. Both these results are in stark contrast to the setting without auxiliary variables where all width measures we consider here differ by more than constant factors and in many cases even by linear factors.
https://arxiv.org/abs/1905.05290
This paper presents an omnidirectional aerial manipulation platform for robust and responsive interaction with unstructured environments, toward the goal of contact-based inspection. The fully actuated tilt-rotor aerial system is equipped with a rigidly mounted end-effector, and is able to exert a 6 degree of freedom force and torque, decoupling the system’s translational and rotational dynamics, and enabling precise interaction with the environment while maintaining stability. An impedance controller with selective apparent inertia is formulated to permit compliance in certain degrees of freedom while achieving precise trajectory tracking and disturbance rejection in others. Experiments demonstrate disturbance rejection, push-and-slide interaction, and on-board state estimation with depth servoing to interact with local surfaces. The system is also validated as a tool for contact-based non-destructive testing of concrete infrastructure.
http://arxiv.org/abs/1905.03502
Pretraining reinforcement learning methods with demonstrations has been an important concept in the study of reinforcement learning since a large amount of computing power is spent on online simulations with existing reinforcement learning algorithms. Pretraining reinforcement learning remains a significant challenge in exploiting expert demonstrations whilst keeping exploration potentials, especially for value based methods. In this paper, we propose a pretraining method for soft Q-learning. Our work is inspired by pretraining methods for actor-critic algorithms since soft Q-learning is a value based algorithm that is equivalent to policy gradient. The proposed method is based on $\gamma$-discounted biased policy evaluation with entropy regularization, which is also the updating target of soft Q-learning. Our method is evaluated on various tasks from Atari 2600. Experiments show that our method effectively learns from imperfect demonstrations, and outperforms other state-of-the-art methods that learn from expert demonstrations.
http://arxiv.org/abs/1905.03501
Significant performance degradation of automatic speech recognition (ASR) systems is observed when the audio signal contains cross-talk. One of the recently proposed approaches to solve the problem of multi-speaker ASR is the deep clustering (DPCL) approach. Combining DPCL with a state-of-the-art hybrid acoustic model, we obtain a word error rate (WER) of 16.5 % on the commonly used wsj0-2mix dataset, which is the best performance reported thus far to the best of our knowledge. The wsj0-2mix dataset contains simulated cross-talk where the speech of multiple speakers overlaps for almost the entire utterance. In a more realistic ASR scenario the audio signal contains significant portions of single-speaker speech and only part of the signal contains speech of multiple competing speakers. This paper investigates obstacles of applying DPCL as a preprocessing method for ASR in such a scenario of sparsely overlapping speech. To this end we present a data simulation approach, closely related to the wsj0-2mix dataset, generating sparsely overlapping speech datasets of arbitrary overlap ratio. The analysis of applying DPCL to sparsely overlapping speech is an important interim step between the fully overlapping datasets like wsj0-2mix and more realistic ASR datasets, such as CHiME-5 or AMI.
http://arxiv.org/abs/1905.03500
The applicability of computer vision to real paintings and artworks has been rarely investigated, even though a vast heritage would greatly benefit from techniques which can understand and process data from the artistic domain. This is partially due to the small amount of annotated artistic data, which is not even comparable to that of natural images captured by cameras. In this paper, we propose a semantic-aware architecture which can translate artworks to photo-realistic visualizations, thus reducing the gap between visual features of artistic and realistic data. Our architecture can generate natural images by retrieving and learning details from real photos through a similarity matching strategy which leverages a weakly-supervised semantic understanding of the scene. Experimental results show that the proposed technique leads to increased realism and to a reduction in domain shift, which improves the performance of pre-trained architectures for classification, detection, and segmentation. Code is publicly available at: https://github.com/aimagelab/art2real.
http://arxiv.org/abs/1811.10666
We propose an efficient two-layer near-lossless coding method using an extended histogram packing technique with backward compatibility to the legacy JPEG standard. The JPEG XT, which is the international standard to compress HDR images, adopts a two-layer coding method for backward compatibility to the legacy JPEG standard. However, there are two problems with this two-layer coding method. One is that it does not exhibit better near-lossless performance than other methods for HDR image compression with single-layer structure. The other problem is that the determining the appropriate values of the coding parameters may be required for each input image to achieve good compression performance of near-lossless compression with the two-layer coding method of the JPEG XT. To solve these problems, we focus on a histogram-packing technique that takes into account the histogram sparseness of HDR images. We used zero-skip quantization, which is an extension of the histogram-packing technique proposed for lossless coding, for implementing the proposed near-lossless coding method. The experimental results indicate that the proposed method exhibits not only a better near-lossless compression performance than that of the two-layer coding method of the JPEG XT, but also there are no issue regarding the combination of parameter values without losing backward compatibility to the JPEG standard.
http://arxiv.org/abs/1905.04129
Packet routing is one of the fundamental problems in computer networks in which a router determines the next-hop of each packet in the queue to get it as quickly as possible to its destination. Reinforcement learning has been introduced to design the autonomous packet routing policy namely Q-routing only using local information available to each router. However, the curse of dimensionality of Q-routing prohibits the more comprehensive representation of dynamic network states, thus limiting the potential benefit of reinforcement learning. Inspired by recent success of deep reinforcement learning (DRL), we embed deep neural networks in multi-agent Q-routing. Each router possesses an independent neural network that is trained without communicating with its neighbors and makes decision locally. Two multi-agent DRL-enabled routing algorithms are proposed: one simply replaces Q-table of vanilla Q-routing by a deep neural network, and the other further employs extra information including the past actions and the destinations of non-head of line packets. Our simulation manifests that the direct substitution of Q-table by a deep neural network may not yield minimal delivery delays because the neural network does not learn more from the same input. When more information is utilized, adaptive routing policy can converge and significantly reduce the packet delivery time.
http://arxiv.org/abs/1905.03494
Deepfake detection is formulated as a hypothesis testing problem to classify an image as genuine or GAN-generated. A robust statistics view of GANs is considered to bound the error probability for various GAN implementations in terms of their performance. The bounds are further simplified using a Euclidean approximation for the low error regime. Lastly, relationships between error probability and epidemic thresholds for spreading processes in networks are established.
http://arxiv.org/abs/1905.03493
Facial landmark localization is a very crucial step in numerous face related applications, such as face recognition, facial pose estimation, face image synthesis, etc. However, previous competitions on facial landmark localization (i.e., the 300-W, 300-VW and Menpo challenges) aim to predict 68-point landmarks, which are incompetent to depict the structure of facial components. In order to overcome this problem, we construct a challenging dataset, named JD-landmark. Each image is manually annotated with 106-point landmarks. This dataset covers large variations on pose and expression, which brings a lot of difficulties to predict accurate landmarks. We hold a 106-point facial landmark localization competition1 on this dataset in conjunction with IEEE International Conference on Multimedia and Expo (ICME) 2019. The purpose of this competition is to discover effective and robust facial landmark localization approaches.
http://arxiv.org/abs/1905.03469
Multi-person pose estimation is an important but challenging problem in computer vision. Although current approaches have achieved significant progress by fusing the multi-scale feature maps, they pay little attention to enhancing the channel-wise and spatial information of the feature maps. In this paper, we propose two novel modules to perform the enhancement of the information for the multi-person pose estimation. First, a Channel Shuffle Module (CSM) is proposed to adopt the channel shuffle operation on the feature maps with different levels, promoting cross-channel information communication among the pyramid feature maps. Second, a Spatial, Channel-wise Attention Residual Bottleneck (SCARB) is designed to boost the original residual unit with attention mechanism, adaptively highlighting the information of the feature maps both in the spatial and channel-wise context. The effectiveness of our proposed modules is evaluated on the COCO keypoint benchmark, and experimental results show that our approach achieves the state-of-the-art results.
http://arxiv.org/abs/1905.03466
Due to the high storage and search efficiency, hashing has become prevalent for large-scale similarity search. Particularly, deep hashing methods have greatly improved the search performance under supervised scenarios. In contrast, unsupervised deep hashing models can hardly achieve satisfactory performance due to the lack of reliable supervisory similarity signals. To address this issue, we propose a novel deep unsupervised hashing model, dubbed DistillHash, which can learn a distilled data set consisted of data pairs, which have confidence similarity signals. Specifically, we investigate the relationship between the initial noisy similarity signals learned from local structures and the semantic similarity labels assigned by a Bayes optimal classifier. We show that under a mild assumption, some data pairs, of which labels are consistent with those assigned by the Bayes optimal classifier, can be potentially distilled. Inspired by this fact, we design a simple yet effective strategy to distill data pairs automatically and further adopt a Bayesian learning framework to learn hash functions from the distilled data set. Extensive experimental results on three widely used benchmark datasets show that the proposed DistillHash consistently accomplishes the state-of-the-art search performance.
http://arxiv.org/abs/1905.03465
In this paper, we develop a stochastic approximation type algorithm to solve finite state/action, infinite-horizon, risk-aware Markov decision processes. Our algorithm has two loops. The inner loop computes the risk by solving a stochastic saddle-point problem. We show that several widely investigated risk measures (e.g. conditional value-at-risk, optimized certainty equivalent, and absolute semi-deviation) can be expressed as stochastic saddle-point problems. The outer loop does $Q-$learning to compute an optimal risk-aware policy. We establish the almost sure convergence and convergence rate of our overall algorithm. For an error tolerance $\epsilon>0$ and learning rate $k\in(1/2,\,1]$, the overall convergence rate of our algorithm is $\Omega((\ln(1/\delta\epsilon)/\epsilon^{2})^{1/k}+(\ln(1/\epsilon))^{1/(1-k)})$ with probability at least $1-\delta$
http://arxiv.org/abs/1805.04238
Early detection of lung cancer is an effective way to improve the survival rate of patients. It is a critical step to have accurate detection of lung nodules in computed tomography (CT) images for the diagnosis of lung cancer. However, due to the heterogeneity of the lung nodules and the complexity of the surrounding environment, robust nodule detection has been a challenging task. In this study, we propose a two-stage convolutional neural network (TSCNN) architecture for lung nodule detection. The CNN architecture in the first stage is based on the improved UNet segmentation network to establish an initial detection of lung nodules. Simultaneously, in order to obtain a high recall rate without introducing excessive false positive nodules, we propose a novel sampling strategy, and use the offline hard mining idea for training and prediction according to the proposed cascaded prediction method. The CNN architecture in the second stage is based on the proposed dual pooling structure, which is built into three 3D CNN classification networks for false positive reduction. Since the network training requires a significant amount of training data, we adopt a data augmentation method based on random mask. Furthermore, we have improved the generalization ability of the false positive reduction model by means of ensemble learning. The proposed method has been experimentally verified on the LUNA dataset. Experimental results show that the proposed TSCNN architecture can obtain competitive detection performance.
http://arxiv.org/abs/1905.03445