3D point cloud - a new signal representation of volumetric objects - is a discrete collection of triples marking exterior object surface locations in 3D space. Conventional imperfect acquisition processes of 3D point cloud - e.g., stereo-matching from multiple viewpoint images or depth data acquired directly from active light sensors - imply non-negligible noise in the data. In this paper, we adopt a previously proposed low-dimensional manifold model for the surface patches in the point cloud and seek self-similar patches to denoise them simultaneously using the patch manifold prior. Due to discrete observations of the patches on the manifold, we approximate the manifold dimension computation defined in the continuous domain with a patch-based graph Laplacian regularizer and propose a new discrete patch distance measure to quantify the similarity between two same-sized surface patches for graph construction that is robust to noise. We show that our graph Laplacian regularizer has a natural graph spectral interpretation, and has desirable numerical stability properties via eigenanalysis. Extensive simulation results show that our proposed denoising scheme can outperform state-of-the-art methods in objective metrics and can better preserve visually salient structural features like edges.
http://arxiv.org/abs/1803.07252
The widespread popularization of vehicles has facilitated all people’s life during the last decades. However, the emergence of a large number of vehicles poses the critical but challenging problem of vehicle re-identification (reID). Till now, for most vehicle reID algorithms, both the training and testing processes are conducted on the same annotated datasets under supervision. However, even a well-trained model will still cause fateful performance drop due to the severe domain bias between the trained dataset and the real-world scenes. To address this problem, this paper proposes a domain adaptation framework for vehicle reID (DAVR), which narrows the cross-domain bias by fully exploiting the labeled data from the source domain to adapt the target domain. DAVR develops an image-to-image translation network named Dual-branch Adversarial Network (DAN), which could promote the images from the source domain (well-labeled) to learn the style of target domain (unlabeled) without any annotation and preserve identity information from source domain. Then the generated images are employed to train the vehicle reID model by a proposed attention-based feature learning model with more reasonable styles. Through the proposed framework, the well-trained reID model has better domain adaptation ability for various scenes in real-world situations. Comprehensive experimental results have demonstrated that our proposed DAVR can achieve excellent performances on both VehicleID dataset and VeRi-776 dataset.
http://arxiv.org/abs/1905.00006
Using joint actuators to drive the skeletal movements is a common practice in character animation, but the resultant torque patterns are often unnatural or infeasible for real humans to achieve. On the other hand, physiologically-based models explicitly simulate muscles and tendons and thus produce more human-like movements and torque patterns. This paper introduces a technique to transform an optimal control problem formulated in the muscle-actuation space to an equivalent problem in the joint-actuation space, such that the solutions to both problems have the same optimal value. By solving the equivalent problem in the joint-actuation space, we can generate human-like motions comparable to those generated by musculotendon models, while retaining the benefit of simple modeling and fast computation offered by joint-actuation models. Our method transforms constant bounds on muscle activations to nonlinear, state-dependent torque limits in the joint-actuation space. In addition, the metabolic energy function on muscle activations is transformed to a nonlinear function of joint torques, joint configuration and joint velocity. Our technique can also benefit policy optimization using deep reinforcement learning approach, by providing a more anatomically realistic action space for the agent to explore during the learning process. We take the advantage of the physiologically-based simulator, OpenSim, to provide training data for learning the torque limits and the metabolic energy function. Once trained, the same torque limits and the energy function can be applied to drastically different motor tasks formulated as either trajectory optimization or policy learning.
http://arxiv.org/abs/1904.13041
Data for face analysis often exhibit highly-skewed class distribution, i.e., most data belong to a few majority classes, while the minority classes only contain a scarce amount of instances. To mitigate this issue, contemporary deep learning methods typically follow classic strategies such as class re-sampling or cost-sensitive training. In this paper, we conduct extensive and systematic experiments to validate the effectiveness of these classic schemes for representation learning on class-imbalanced data. We further demonstrate that more discriminative deep representation can be learned by enforcing a deep network to maintain inter-cluster margins both within and between classes. This tight constraint effectively reduces the class imbalance inherent in the local data neighborhood, thus carving much more balanced class boundaries locally. We show that it is easy to deploy angular margins between the cluster distributions on a hypersphere manifold. Such learned Cluster-based Large Margin Local Embedding (CLMLE), when combined with a simple k-nearest cluster algorithm, shows significant improvements in accuracy over existing methods on both face recognition and face attribute prediction tasks that exhibit imbalanced class distribution.
http://arxiv.org/abs/1806.00194
This paper presents a wearable assistive device with the shape of a pair of eyeglasses that allows visually impaired people to navigate safely and quickly in unfamiliar environment, as well as perceive the complicated environment to automatically make decisions on the direction to move. The device uses a consumer Red, Green, Blue and Depth (RGB-D) camera and an Inertial Measurement Unit (IMU) to detect obstacles. As the device leverages the ground height continuity among adjacent image frames, it is able to segment the ground from obstacles accurately and rapidly. Based on the detected ground, the optimal walkable direction is computed and the user is then informed via converted beep sound. Moreover, by utilizing deep learning techniques, the device can semantically categorize the detected obstacles to improve the users’ perception of surroundings. It combines a Convolutional Neural Network (CNN) deployed on a smartphone with a depth-image-based object detection to decide what the object type is and where the object is located, and then notifies the user of such information via speech. We evaluated the device’s performance with different experiments in which 20 visually impaired people were asked to wear the device and move in an office, and found that they were able to avoid obstacle collisions and find the way in complicated scenarios.
http://arxiv.org/abs/1904.13037
This paper presents a novel garbage pickup robot which operates on the grass. The robot is able to detect the garbage accurately and autonomously by using a deep neural network for garbage recognition. In addition, with the ground segmentation using a deep neural network, a novel navigation strategy is proposed to guide the robot to move around. With the garbage recognition and automatic navigation functions, the robot can clean garbage on the ground in places like parks or schools efficiently and autonomously. Experimental results show that the garbage recognition accuracy can reach as high as 95%, and even without path planning, the navigation strategy can reach almost the same cleaning efficiency with traditional methods. Thus, the proposed robot can serve as a good assistance to relieve dustman’s physical labor on garbage cleaning tasks.
http://arxiv.org/abs/1904.13034
Self-driving industry vehicle plays a key role in the industry automation and contributes to resolve the problems of the shortage and increasing cost in manpower. Place recognition and loop-closure detection are main challenges in the localization and navigation tasks, specially when industry vehicles work in large-scale complex environments, such as the logistics warehouse and the port terminal. In this paper, we resolve the loop-closure detection problem by developing a novel 3D point cloud learning network, an active super keyframe selection method and a coarse-to-fine sequence matching strategy. More specifically, we first propose a novel deep neural network to extract a global descriptors from the original large-scale 3D point cloud, then based on which, an environment analysis approach is presented to investigate the feature space distribution of the global descriptors and actively select several super keyframes. Finally, a coarse-to-fine sequence matching strategy, which includes a super keyframe based coarse matching stage and a local sequence matching stage, is presented to ensure the loop-closure detection accuracy and real-time performance simultaneously. The proposed network is evaluated in different datasets and obtains a substantial improvement against the state-of-the-art PointNetVLAD in place recognition tasks. Experiment results on a self-driving industry vehicle validate the effectiveness of the proposed loop-closure detection algorithm.
http://arxiv.org/abs/1904.13030
To help the blind people walk to the destination efficiently and safely in indoor environment, a novel wearable navigation device is presented in this paper. The locating, way-finding, route following and obstacle avoiding modules are the essential components in a navigation system, while it remains a challenging task to consider obstacle avoiding during route following, as the indoor environment is complex, changeable and possibly with dynamic objects. To address this issue, we propose a novel scheme which utilizes a dynamic sub-goal selecting strategy to guide the users to the destination and help them bypass obstacles at the same time. This scheme serves as the key component of a complete navigation system deployed on a pair of wearable optical see-through glasses for the ease of use of blind people’s daily walks. The proposed navigation device has been tested on a collection of individuals and proved to be effective on indoor navigation tasks. The sensors embedded are of low cost, small volume and easy integration, making it possible for the glasses to be widely used as a wearable consumer device.
http://arxiv.org/abs/1904.13028
In radiology, radiologists not only detect lesions from the medical image, but also describe them with various attributes such as their type, location, size, shape, and intensity. While these lesion attributes are rich and useful in many downstream clinical applications, how to extract them from the radiology reports is less studied. This paper outlines a novel deep learning method to automatically extract attributes of lesions of interest from the clinical text. Different from classical CNN models, we integrated the multi-head self-attention mechanism to handle the long-distance information in the sentence, and to jointly correlate different portions of sentence representation subspaces in parallel. Evaluation on an in-house corpus demonstrates that our method can achieve high performance with 0.848 in precision, 0.788 in recall, and 0.815 in F-score. The new method and constructed corpus will enable us to build automatic systems with a higher-level understanding of the radiological world.
http://arxiv.org/abs/1904.13018
Encoder-decoder based neural architectures serve as the basis of state-of-the-art approaches in end-to-end open domain dialog systems. Since most of such systems are trained with a maximum likelihood(MLE) objective they suffer from issues such as lack of generalizability and the generic response problem, i.e., a system response that can be an answer to a large number of user utterances, e.g., “Maybe, I don’t know.” Having explicit feedback on the relevance and interestingness of a system response at each turn can be a useful signal for mitigating such issues and improving system quality by selecting responses from different approaches. Towards this goal, we present a system that evaluates chatbot responses at each dialog turn for coherence and engagement. Our system provides explicit turn-level dialog quality feedback, which we show to be highly correlated with human evaluation. To show that incorporating this feedback in the neural response generation models improves dialog quality, we present two different and complementary mechanisms to incorporate explicit feedback into a neural response generation model: reranking and direct modification of the loss function during training. Our studies show that a response generation model that incorporates these combined feedback mechanisms produce more engaging and coherent responses in an open-domain spoken dialog setting, significantly improving the response quality using both automatic and human evaluation.
http://arxiv.org/abs/1904.13015
We evaluate machine learning methods for event classification in the Active-Target Time Projection Chamber detector at the National Superconducting Cyclotron Laboratory (NSCL) at Michigan State University. An automated method to single out the desired reaction product would result in more accurate physics results as well as a faster analysis process. Binary and multi-class classification methods were tested on data produced by the $^{46}$Ar(p,p) experiment run at the NSCL in September 2015. We found a Convolutional Neural Network to be the most successful classifier of proton scattering events for transfer learning. Results from this investigation and recommendations for event classification in future experiments are presented.
http://arxiv.org/abs/1810.10350
In order to solve complex, long-horizon tasks, intelligent robots need to be able to carry out high-level, abstract planning and reasoning in conjunction with motion planning. However, abstract models are typically lossy and plans or policies computed using them are often unexecutable in practice. These problems are aggravated in more realistic situations with stochastic dynamics, where the robot needs to reason about, and plan for multiple possible contingencies. We present a new approach for integrated task and motion planning in such settings. In contrast to prior work in this direction, we show that our approach can effectively compute integrated task and motion policies with branching structure encoding agent behaviors for various possible contingencies. We prove that our algorithm is probabilistically complete and can compute feasible solution policies in an anytime fashion so that the probability of encountering an unresolved contingency decreases over time. Empirical results on a set of challenging problems show the utility and scope of our methods.
http://arxiv.org/abs/1904.13006
In this paper, a novel signature of human action recognition, namely the curvature of a video sequence, is introduced. In this way, the distribution of sequential data is modeled, which enables few-shot learning. Instead of depending on recognizing features within images, our algorithm views actions as sequences on the universal time scale across a whole sequence of images. The video sequence, viewed as a curve in pixel space, is aligned by reparameterization using the arclength of the curve in pixel space. Once such curvatures are obtained, statistical indexes are extracted and fed into a learning-based classifier. Overall, our method is simple but powerful. Preliminary experimental results show that our method is effective and achieves state-of-the-art performance in video-based human action recognition. Moreover, we see latent capacity in transferring this idea into other sequence-based recognition applications such as speech recognition, machine translation, and text generation.
http://arxiv.org/abs/1904.13003
The recent introduction of the AVA dataset for action detection has caused a renewed interest to this problem. Several approaches have been recently proposed that improved the performance. However, all of them have ignored the main difficulty of the AVA dataset - its realistic distribution of training and test examples. This dataset was collected by exhaustive annotation of human action in uncurated videos. As a result, the most common categories, such as stand' or sit’, contain tens of thousands of examples, where rare ones have only dozens. In this work we study the problem of action detection in highly-imbalanced dataset. Differently from previous work on handling long-tail category distributions, we begin by analyzing the imbalance in the test set. We demonstrate that the standard AP metric is not informative for the categories in the tail, and propose an alternative one - Sampled AP. Armed with this new measure, we study the problem of transferring representations from the data-rich head to the rare tail categories and propose a simple but effective approach.
http://arxiv.org/abs/1904.12993
Methods for interpreting machine learning black-box models increase the outcomes’ transparency and in turn generates insight into the reliability and fairness of the algorithms. However, the interpretations themselves could contain significant uncertainty that undermines the trust in the outcomes and raises concern about the model’s reliability. Focusing on the method “Local Interpretable Model-agnostic Explanations” (LIME), we demonstrate the presence of two sources of uncertainty, namely the randomness in its sampling procedure and the variation of interpretation quality across different input data points. Such uncertainty is present even in models with high training and test accuracy. We apply LIME to synthetic data and two public data sets, text classification in 20 Newsgroup and recidivism risk-scoring in COMPAS, to support our argument.
http://arxiv.org/abs/1904.12991
Computing similarity between a query and a document is fundamental in any information retrieval system. In search engines, computing query-document similarity is an essential step in both retrieval and ranking stages. In eBay search, document is an item and the query-item similarity can be computed by comparing different facets of the query-item pair. Query text can be compared with the text of the item title. Likewise, a category constraint applied on the query can be compared with the listing category of the item. However, images are one signal that are usually present in the items but are not present in the query. Images are one of the most intuitive signals used by users to determine the relevance of the item given a query. Including this signal in estimating similarity between the query-item pair is likely to improve the relevance of the search engine. We propose a novel way of deriving image information for queries. We attempt to learn image information for queries from item images instead of generating explicit image features or an image for queries. We use canonical correlation analysis (CCA) to learn a new subspace where projecting the original data will give us a new query and item representation. We hypothesize that this new query representation will also have image information about the query. We estimate the query-item similarity using a vector space model and report the performance of the proposed method on eBay’s search data. We show 11.89\% relevance improvement over the baseline using area under the receiver operating characteristic curve (AUROC) as the evaluation metric. We also show 3.1\% relevance improvement over the baseline with area under the precision recall curve (AUPRC) .
http://arxiv.org/abs/1904.12856
In this paper we present a generic and query-efficient black-box attack demonstrated against API call based machine learning malware classifiers. We generate adversarial examples combining sequences (API call sequences) and other features (e.g., printable strings) that will be misclassified by the classifier without affecting the malware functionality. Opposed to previous studies, our attack minimizes the number of target recurrent neural network (RNN) classifier queries and only requires access to the predicted label of the attacked model (with or without its confidence score). We evaluate the attack’s effectiveness against a variety of classifiers, including RNN variants, deep neural networks, support vector machines, and gradient-boosted decision trees. Our attack success rate is about 98% when the confidence score is known and 88% when only the label is known. We implement four state of the art query-efficient attacks for sequence input and show that our attack requires fewer queries and less knowledge about the attacked model’s architecture than other existing query-efficient attacks, making it practical for attacking cloud based models at a minimal cost.
https://arxiv.org/abs/1804.08778
Generating training sets for deep convolutional neural networks (DCNNs) is a bottleneck for modern real-world applications. This is a demanding task for applications where annotating training data is costly, such as in semantic segmentation. In the literature, there is still a gap between the performance achieved by a network trained on full and on weak annotations. In this paper, we establish a strategy to measure this gap and to identify the ingredients necessary to reduce it. On scribbles, we establish new state-of-the-art results: we obtain a mIoU of 75.6% without, and 75.7% with CRF post-processing. We reduce the gap by 64.2% whereas the current state-of-the-art reduces it only by 57.5%. Thanks to a systematic study of the different ingredients involved in the weakly supervised scenario and an original experimental strategy, we unravel a counter-intuitive mechanism that is simple and amenable to generalisations to other weakly-supervised scenarios: averaging poor local predicted annotations with the baseline ones and reuse them for training a DCNN yields new state-of-the-art results.
http://arxiv.org/abs/1808.01625
The recent adoption of Electronic Health Records (EHRs) by health care providers has introduced an important source of data that provides detailed and highly specific insights into patient phenotypes over large cohorts. These datasets, in combination with machine learning and statistical approaches, generate new opportunities for research and clinical care. However, many methods require the patient representations to be in structured formats, while the information in the EHR is often locked in unstructured texts designed for human readability. In this work, we develop the methodology to automatically extract clinical features from clinical narratives from large EHR corpora without the need for prior knowledge. We consider medical terms and sentences appearing in clinical narratives as atomic information units. We propose an efficient clustering strategy suitable for the analysis of large text corpora and to utilize the clusters to represent information about the patient compactly. To demonstrate the utility of our approach, we perform an association study of clinical features with somatic mutation profiles from 4,007 cancer patients and their tumors. We apply the proposed algorithm to a dataset consisting of about 65 thousand documents with a total of about 3.2 million sentences. We identify 341 significant statistical associations between the presence of somatic mutations and clinical features. We annotated these associations according to their novelty, and report several known associations. We also propose 32 testable hypotheses where the underlying biological mechanism does not appear to be known but plausible. These results illustrate that the automated discovery of clinical features is possible and the joint analysis of clinical and genetic datasets can generate appealing new hypotheses.
http://arxiv.org/abs/1904.12973
This paper proposes a new neural network based on SPD manifold learning for skeleton-based hand gesture recognition. Given the stream of hand’s joint positions, our approach combines two aggregation processes on respectively spatial and temporal domains. The pipeline of our network architecture consists in three main stages. The first stage is based on a convolutional layer to increase the discriminative power of learned features. The second stage relies on different architectures for spatial and temporal Gaussian aggregation of joint features. The third stage learns a final SPD matrix from skeletal data. A new type of layer is proposed for the third stage, based on a variant of stochastic gradient descent on Stiefel manifolds. The proposed network is validated on two challenging datasets and shows state-of-the-art accuracies on both datasets.
http://arxiv.org/abs/1904.12970
Neural circuits can be reconstructed from brain images acquired by serial section electron microscopy. Image analysis has been performed by manual labor for half a century, and efforts at automation date back almost as far. Convolutional nets were first applied to neuronal boundary detection a dozen years ago, and have now achieved impressive accuracy on clean images. Robust handling of image defects is a major outstanding challenge. Convolutional nets are also being employed for other tasks in neural circuit reconstruction: finding synapses and identifying synaptic partners, extending or pruning neuronal reconstructions, and aligning serial section images to create a 3D image stack. Computational systems are being engineered to handle petavoxel images of cubic millimeter brain volumes.
http://arxiv.org/abs/1904.12966
The humanity has been facing a plethora of challenges associated with infectious diseases, which kill more than 6 million people a year. Although continuous efforts have been applied to relieve the potential damages from such misfortunate events, it is unquestionable that there are many persisting challenges yet to overcome. One related issue we particularly address here is the assessment and prediction of such epidemics. In this field of study, traditional and ad-hoc models frequently fail to provide proper predictive situation awareness (PSAW), characterized by understanding the current situations and predicting the future situations. Comprehensive PSAW for infectious disease can support decision making and help to hinder disease spread. In this paper, we develop a computing system platform focusing on collective intelligence causal modeling, in order to support PSAW in the domain of infectious disease. Analyses of global epidemics require integration of multiple different data and models, which can be originated from multiple independent researchers. These models should be integrated to accurately assess and predict the infectious disease in terms of holistic view. The system shall provide three main functions: (1) collaborative causal modeling, (2) causal model integration, and (3) causal model reasoning. These functions are supported by subject-matter expert and artificial intelligence (AI), with uncertainty treatment. Subject-matter experts, as collective intelligence, develop causal models and integrate them as one joint causal model. The integrated causal model shall be used to reason about: (1) the past, regarding how the causal factors have occurred; (2) the present, regarding how the spread is going now; and (3) the future, regarding how it will proceed. Finally, we introduce one use case of predictive situation awareness for the Ebola virus disease.
http://arxiv.org/abs/1904.12958
In this paper, we present new data pre-processing and augmentation techniques for DNN-based raw image denoising. Compared with traditional RGB image denoising, performing this task on direct camera sensor readings presents new challenges such as how to effectively handle various Bayer patterns from different data sources, and subsequently how to perform valid data augmentation with raw images. To address the first problem, we propose a Bayer pattern unification (BayerUnify) method to unify different Bayer patterns. This allows us to fully utilize a heterogeneous dataset to train a single denoising model instead of training one model for each pattern. Furthermore, while it is essential to augment the dataset to improve model generalization and performance, we discovered that it is error-prone to modify raw images by adapting augmentation methods designed for RGB images. Towards this end, we present a Bayer preserving augmentation (BayerAug) method as an effective approach for raw image augmentation. Combining these data processing technqiues with a modified U-Net, our method achieves a PSNR of 52.11 and a SSIM of 0.9969 in NTIRE 2019 Real Image Denoising Challenge, demonstrating the state-of-the-art performance.
http://arxiv.org/abs/1904.12945
In bipedal gait design literature, one of the common ways of generating stable 3D walking gait is by designing the frontal and sagittal controllers as decoupled dynamics. The study of the decoupled frontal dynamics is, however, still understudied if compared with the sagittal dynamics. In this paper it is presented a formal approach to the problem of frontal dynamics stabilization by extending the hybrid zero dynamics framework to deal with the frontal gait design problem.
http://arxiv.org/abs/1904.12939
The Bayesian brain hypothesis, predictive processing and variational free energy minimisation are typically used to describe perceptual processes based on accurate generative models of the world. However, generative models need not be veridical representations of the environment. We suggest that they can (and should) be used to describe sensorimotor relationships relevant for behaviour rather than precise accounts of the world.
http://arxiv.org/abs/1904.12937
We address the problem of finding a set of images containing a common, but unknown, object category from a collection of image proposals. Our formulation assumes that we are given a collection of bags where each bag is a set of image proposals. Our goal is to select one image from each bag such that the selected images are of the same object category. We model the selection as an energy minimization problem with unary and pairwise potential functions. Inspired by recent few-shot learning algorithms, we propose an approach to learn the potential functions directly from the data. Furthermore, we propose a fast and simple greedy inference algorithm for energy minimization. We evaluate our approach on few-shot common object recognition and object co-localization tasks. Our experiments show that learning the pairwise and unary terms greatly improves the performance of the model over several well-known methods for these tasks. The proposed greedy optimization algorithm achieves performance comparable to state-of-the-art structured inference algorithms while being ~10 times faster. The code is publicly available on https://github.com/haamoon/finding_common_object.
http://arxiv.org/abs/1904.12936
This paper presents our work of training acoustic event detection (AED) models using unlabeled dataset. Recent acoustic event detectors are based on large-scale neural networks, which are typically trained with huge amounts of labeled data. Labels for acoustic events are expensive to obtain, and relevant acoustic event audios can be limited, especially for rare events. In this paper we leverage an Internet-scale unlabeled dataset with potential domain shift to improve the detection of acoustic events. Based on the classic tri-training approach, our proposed method shows accuracy improvement over both the supervised training baseline, and semisupervised self-training set-up, in all pre-defined acoustic event detection tasks. As our approach relies on ensemble models, we further show the improvements can be distilled to a single model via knowledge distillation, with the resulting single student model maintaining high accuracy of teacher ensemble models.
http://arxiv.org/abs/1904.12926
Enabling robots to understand instructions provided via spoken natural language would facilitate interaction between robots and people in a variety of settings in homes and workplaces. However, natural language instructions are often missing information that would be obvious to a human based on environmental context and common sense, and hence does not need to be explicitly stated. In this paper, we introduce Language-Model-based Commonsense Reasoning (LMCR), a new method which enables a robot to listen to a natural language instruction from a human, observe the environment around it, and automatically fill in information missing from the instruction using environmental context and a new commonsense reasoning approach. Our approach first converts an instruction provided as unconstrained natural language into a form that a robot can understand by parsing it into verb frames. Our approach then fills in missing information in the instruction by observing objects in its vicinity and leveraging commonsense reasoning. To learn commonsense reasoning automatically, our approach distills knowledge from large unstructured textual corpora by training a language model. Our results show the feasibility of a robot learning commonsense knowledge automatically from web-based textual corpora, and the power of learned commonsense reasoning models in enabling a robot to autonomously perform tasks based on incomplete natural language instructions.
http://arxiv.org/abs/1904.12907
Reinforcement learning (RL) has proven its worth in a series of artificial domains, and is beginning to show some successes in real-world scenarios. However, much of the research advances in RL are often hard to leverage in real-world systems due to a series of assumptions that are rarely satisfied in practice. We present a set of nine unique challenges that must be addressed to productionize RL to real world problems. For each of these challenges, we specify the exact meaning of the challenge, present some approaches from the literature, and specify some metrics for evaluating that challenge. An approach that addresses all nine challenges would be applicable to a large number of real world problems. We also present an example domain that has been modified to present these challenges as a testbed for practical RL research.
http://arxiv.org/abs/1904.12901
Recent studies on medical image synthesis reported promising results using generative adversarial networks, mostly focusing on one-to-one cross-modality synthesis. Naturally, the idea arises that a target modality would benefit from multi-modal input. Synthesizing MR imaging sequences is highly attractive for clinical practice, as often single sequences are missing or of poor quality (e.g. due to motion). However, existing methods fail to scale up to image volumes with high numbers of modalities and extensive non-aligned volumes, facing common draw-backs of complex multi-modal imaging sequences. To address these limitations, we propose a novel, scalable and multi-modal approach calledDiamondGAN. Our model is capable of performing flexible non-aligned cross-modality synthesis and data infill, when given multiple modalities or any of their arbitrary subsets. It learns structured information using non-aligned input modalities in an end-to-end fashion. We synthesize two MRI sequences with clinical relevance (i.e., double inversion recovery (DIR) and contrast-enhanced T1 (T1-c)), which are reconstructed from three common MRI sequences. In addition, we perform multi-rater visual evaluation experiment and find that trained radiologists are unable to distinguish our synthetic DIR images from real ones.
http://arxiv.org/abs/1904.12894
Deep generative architectures provide a way to model not only images but also complex, 3-dimensional objects, such as point clouds. In this work, we present a novel method to obtain meaningful representations of 3D shapes that can be used for challenging tasks including 3D points generation, reconstruction, compression, and clustering. Contrary to existing methods for 3D point cloud generation that train separate decoupled models for representation learning and generation, our approach is the first end-to-end solution that allows to simultaneously learn a latent space of representation and generate 3D shape out of it. Moreover, our model is capable of learning meaningful compact binary descriptors with adversarial training conducted on a latent space. To achieve this goal, we extend a deep Adversarial Autoencoder model (AAE) to accept 3D input and create 3D output. Thanks to our end-to-end training regime, the resulting method called 3D Adversarial Autoencoder (3dAAE) obtains either binary or continuous latent space that covers a much wider portion of training data distribution. Finally, our quantitative evaluation shows that 3dAAE provides state-of-the-art results for 3D points clustering and 3D object retrieval.
http://arxiv.org/abs/1811.07605
Despite its success, deep learning still needs large labeled datasets to succeed. Data augmentation has shown much promise in alleviating the need for more labeled data, but it so far has mostly been applied in supervised settings and achieved limited gains. In this work, we propose to apply data augmentation to unlabeled data in a semi-supervised learning setting. Our method, named Unsupervised Data Augmentation or UDA, encourages the model predictions to be consistent between an unlabeled example and an augmented unlabeled example. Unlike previous methods that use random noise such as Gaussian noise or dropout noise, UDA has a small twist in that it makes use of harder and more realistic noise generated by state-of-the-art data augmentation methods. This small twist leads to substantial improvements on six language tasks and three vision tasks even when the labeled set is extremely small. For example, on the IMDb text classification dataset, with only 20 labeled examples, UDA outperforms the state-of-the-art model trained on 25,000 labeled examples. On standard semi-supervised learning benchmarks, CIFAR-10 with 4,000 examples and SVHN with 1,000 examples, UDA outperforms all previous approaches and reduces more than $30\%$ of the error rates of state-of-the-art methods: going from 7.66% to 5.27% and from 3.53% to 2.46% respectively. UDA also works well on datasets that have a lot of labeled data. For example, on ImageNet, with 1.3M extra unlabeled data, UDA improves the top-1/top-5 accuracy from 78.28/94.36% to 79.04/94.45% when compared to AutoAugment.
http://arxiv.org/abs/1904.12848
Adversarial training, in which a network is trained on adversarial examples, is one of the few defenses against adversarial attacks that withstands strong attacks. Unfortunately, the high cost of generating strong adversarial examples makes standard adversarial training impractical on large-scale problems like ImageNet. We present an algorithm that eliminates the overhead cost of generating adversarial examples by recycling the gradient information computed when updating model parameters. Our “free” adversarial training algorithm achieves state-of-the-art robustness on CIFAR-10 and CIFAR-100 datasets at negligible additional cost compared to natural training, and can be 7 to 30 times faster than other strong adversarial training methods. Using a single workstation with 4 P100 GPUs and 2 days of runtime, we can train a robust model for the large-scale ImageNet classification task that maintains 40% accuracy against PGD attacks.
http://arxiv.org/abs/1904.12843
Electric Vehicle (EV) sharing systems have recently experienced unprecedented growth across the globe. During their fast expansion, one fundamental determinant for success is the capability of dynamically predicting the demand of stations as the entire system is evolving continuously. There are several challenges in this dynamic demand prediction problem. Firstly, unlike most of the existing work which predicts demand only for static systems or at few stages of expansion, in the real world we often need to predict the demand as or even before stations are being deployed or closed, to provide information and support for decision making. Secondly, for the stations to be deployed, there is no historical record or additional mobility data available to help the prediction of their demand. Finally, the impact of deploying/closing stations to the remaining stations in the system can be very complex. To address these challenges, in this paper we propose a novel dynamic demand prediction approach based on graph sequence learning, which is able to model the dynamics during the system expansion and predict demand accordingly. We use a local temporal encoding process to handle the available historical data at individual stations, and a dynamic spatial encoding process to take correlations between stations into account with graph convolutional neural networks. The encoded features are fed to a multi-scale prediction network, which forecasts both the long-term expected demand of the stations and their instant demand in the near future. We evaluate the proposed approach on real-world data collected from a major EV sharing platform in Shanghai for one year. Experimental results demonstrate that our approach significantly outperforms the state of the art, showing up to three-fold performance gain in predicting demand for the rapidly expanding EV sharing system.
http://arxiv.org/abs/1903.04051
In this paper we present a self-supervised method for representation learning utilizing two different modalities. Based on the observation that cross-modal information has a high semantic meaning we propose a method to effectively exploit this signal. For our approach we utilize video data since it is available on a large scale and provides easily accessible modalities given by RGB and optical flow. We demonstrate state-of-the-art performance on highly contested action recognition datasets in the context of self-supervised learning. We show that our feature representation also transfers to other tasks and conduct extensive ablation studies to validate our core contributions. Code and model can be found at https://github.com/nawidsayed/Cross-and-Learn.
http://arxiv.org/abs/1811.03879
We present techniques for automatically inferring invariant properties of feed-forward neural networks. Our insight is that feed forward networks should be able to learn a decision logic that is captured in the activation patterns of its neurons. We propose to extract such decision patterns that can be considered as invariants of the network with respect to a certain output behavior. We present techniques to extract input invariants as convex predicates on the input space, and layer invariants that represent features captured in the hidden layers. We apply the techniques on the networks for the MNIST and ACASXU applications. Our experiments highlight the use of invariants in a variety of applications, such as explainability, providing robustness guarantees, detecting adversaries, simplifying proofs and network distillation.
http://arxiv.org/abs/1904.13215
Recent work has raised concerns on the risk of unintended bias in AI systems being used nowadays that can affect individuals unfairly based on race, gender or religion, among other possible characteristics. While a lot of bias metrics and fairness definitions have been proposed in recent years, there is no consensus on which metric/definition should be used and there are very few available resources to operationalize them. Therefore, despite recent awareness, auditing for bias and fairness when developing and deploying AI systems is not yet a standard practice. We present Aequitas, an open source bias and fairness audit toolkit that is an intuitive and easy to use addition to the machine learning workflow, enabling users to seamlessly test models for several bias and fairness metrics in relation to multiple population sub-groups. Aequitas facilitates informed and equitable decisions around developing and deploying algorithmic decision making systems for both data scientists, machine learning researchers and policymakers.
http://arxiv.org/abs/1811.05577
Existing item-based collaborative filtering (ICF) methods leverage only the relation of collaborative similarity. Nevertheless, there exist multiple relations between items in real-world scenarios. Distinct from the collaborative similarity that implies co-interact patterns from the user perspective, these relations reveal fine-grained knowledge on items from different perspectives of meta-data, functionality, etc. However, how to incorporate multiple item relations is less explored in recommendation research. In this work, we propose Relational Collaborative Filtering (RCF), a general framework to exploit multiple relations between items in recommender system. We find that both the relation type and the relation value are crucial in inferring user preference. To this end, we develop a two-level hierarchical attention mechanism to model user preference. The first-level attention discriminates which types of relations are more important, and the second-level attention considers the specific relation values to estimate the contribution of a historical item in recommending the target item. To make the item embeddings be reflective of the relational structure between items, we further formulate a task to preserve the item relations, and jointly train it with the recommendation task of preference modeling. Empirical results on two real datasets demonstrate the strong performance of RCF. Furthermore, we also conduct qualitative analyses to show the benefits of explanations brought by the modeling of multiple item relations.
http://arxiv.org/abs/1904.12796
We tackle the problem of texture synthesis in the setting where many input images are given and a large-scale output is required. We build on recent generative adversarial networks and propose two extensions in this paper. First, we propose an algorithm to combine outputs of GANs trained on a smaller resolution to produce a large-scale plausible texture map with virtually no boundary artifacts. Second, we propose a user interface to enable artistic control. Our quantitative and qualitative results showcase the generation of synthesized high-resolution maps consisting of up to hundreds of megapixels as a case in point.
http://arxiv.org/abs/1904.12795
Style transfer algorithms strive to render the content of one image using the style of another. We propose Style Transfer by Relaxed Optimal Transport and Self-Similarity (STROTSS), a new optimization-based style transfer algorithm. We extend our method to allow user-specified point-to-point or region-to-region control over visual similarity between the style image and the output. Such guidance can be used to either achieve a particular visual effect or correct errors made by unconstrained style transfer. In order to quantitatively compare our method to prior work, we conduct a large-scale user study designed to assess the style-content tradeoff across settings in style transfer algorithms. Our results indicate that for any desired level of content preservation, our method provides higher quality stylization than prior work. Code is available at https://github.com/nkolkin13/STROTSS
http://arxiv.org/abs/1904.12785
In this paper, we study a discrete system of entities residing on a two-dimensional square grid. Each entity is modelled as a node occupying a distinct cell of the grid. The set of all $n$ nodes forms initially a connected shape $A$. Entities are equipped with a linear-strength pushing mechanism that can push a whole line of entities, from 1 to $n$, in parallel in a single time-step. A target connected shape $B$ is also provided and the goal is to \emph{transform} $A$ into $B$ via a sequence of line movements. Existing models based on local movement of individual nodes, such as rotating or sliding a single node, can be shown to be special cases of the present model, therefore their (inefficient, $\Theta(n^2)$) \emph{universal transformations} carry over. Our main goal is to investigate whether the parallelism inherent in this new type of movement can be exploited for efficient, i.e., sub-quadratic worst-case, transformations. As a first step towards this, we restrict attention solely to centralised transformations and leave the distributed case as a direction for future research. Our results are positive. By focusing on the apparently hard instance of transforming a diagonal $A$ into a straight line $B$, we first obtain transformations of time $O(n\sqrt{n})$ without and with preserving the connectivity of the shape throughout the transformation. Then, we further improve by providing two $O(n\log n)$-time transformations for this problem. By building upon these ideas, we first manage to develop an $O(n\sqrt{n})$-time universal transformation. Our main result is then an $ O(n \log n) $-time universal transformation. We leave as an interesting open problem a suspected $\Omega(n\log n)$-time lower bound.
http://arxiv.org/abs/1904.12777
This paper investigates the joint localization, detection, and tracking of sound events using a convolutional recurrent neural network (CRNN). We use a CRNN previously proposed for the localization and detection of stationary sources, and show that the recurrent layers enable the spatial tracking of moving sources when trained with dynamic scenes. The tracking performance of the CRNN is compared with a stand-alone tracking method that combines a multi-source (DOA) estimator and a particle filter. Their respective performance is evaluated in various acoustic conditions such as anechoic and reverberant scenarios, stationary and moving sources at several angular velocities, and with a varying number of overlapping sources. The results show that the CRNN manages to track multiple sources more consistently than the parametric method across acoustic scenarios, but at the cost of higher localization error.
http://arxiv.org/abs/1904.12769
In this paper, we extend the persona-based sequence-to-sequence (Seq2Seq) neural network conversation model to a multi-turn dialogue scenario by modifying the state-of-the-art hredGAN architecture to simultaneously capture utterance attributes such as speaker identity, dialogue topic, speaker sentiments and so on. The proposed system, phredGAN has a persona-based HRED generator (PHRED) and a conditional discriminator. We also explore two approaches to accomplish the conditional discriminator: (1) phredGAN_a, a system that passes the attribute representation as an additional input into a traditional adversarial discriminator, and (2) phredGAN_d, a dual discriminator system which in addition to the adversarial discriminator, collaboratively predicts the attribute(s) that generated the input utterance. To demonstrate the superior performance of phredGAN over the persona Seq2Seq model, we experiment with two conversational datasets, the Ubuntu Dialogue Corpus (UDC) and TV series transcripts from the Big Bang Theory and Friends. Performance comparison is made with respect to a variety of quantitative measures as well as crowd-sourced human evaluation. We also explore the trade-offs from using either variant of phredGAN on datasets with many but weak attribute modalities (such as with Big Bang Theory and Friends) and ones with few but strong attribute modalities (customer-agent interactions in Ubuntu dataset).
http://arxiv.org/abs/1905.01992
In this paper, we extend the persona-based sequence-to-sequence (Seq2Seq) neural network conversation model to multi-turn dialogue by modifying the state-of-the-art hredGAN architecture. To achieve this, we introduce an additional input modality into the encoder and decoder of hredGAN to capture other attributes such as speaker identity, location, sub-topics, and other external attributes that might be available from the corpus of human-to-human interactions. The resulting persona hredGAN ($phredGAN$) shows better performance than both the existing persona-based Seq2Seq and hredGAN models when those external attributes are available in a multi-turn dialogue corpus. This superiority is demonstrated on TV drama series with character consistency (such as Big Bang Theory and Friends) and customer service interaction datasets such as Ubuntu dialogue corpus in terms of perplexity, BLEU, ROUGE, and Distinct n-gram scores.
http://arxiv.org/abs/1905.01998
Recently, differentiable search methods have made major progress in reducing the computational costs of neural architecture search. However, these approaches often report lower accuracy in evaluating the searched architecture or transferring it to another dataset. This is arguably due to the large gap between the architecture depths in search and evaluation scenarios. In this paper, we present an efficient algorithm which allows the depth of searched architectures to grow gradually during the training procedure. This brings two issues, namely, heavier computational overheads and weaker search stability, which we solve using search space approximation and regularization, respectively. With a significantly reduced search time (~7 hours on a single GPU), our approach achieves state-of-the-art performance on both the proxy dataset (CIFAR10 or CIFAR100) and the target dataset (ImageNet). Code is available at https://github.com/chenxin061/pdarts.
http://arxiv.org/abs/1904.12760
We reformulate the option framework as two parallel augmented MDPs. Under this novel formulation, all policy optimization algorithms can be used off the shelf to learn intra-option policies, option termination conditions, and a master policy over options. We apply an actor-critic algorithm on each augmented MDP, yielding the Double Actor-Critic (DAC) architecture. Furthermore, we show that, when state-value functions are used as critics, one critic can be expressed in terms of the other, and hence only one critic is necessary. Our experiments on challenging robot simulation tasks demonstrate that DAC outperforms previous gradient-based option learning algorithms by a large margin and significantly outperforms its hierarchy-free counterparts in a transfer learning setting.
http://arxiv.org/abs/1904.12691
This paper addresses the problem of efficiently computing higher-order variational integrators in simulation and trajectory optimization of mechanical systems as those often found in robotic applications. We develop $O(n)$ algorithms to evaluate the discrete Euler-Lagrange (DEL) equations and compute the Newton direction for solving the DEL equations, which results in linear-time variational integrators of arbitrarily high order. To our knowledge, no linear-time higher-order variational or even implicit integrators have been developed before. Moreover, an $O(n^2)$ algorithm to linearize the DEL equations is presented, which is useful for trajectory optimization. These proposed algorithms eliminate the bottleneck of implementing higher-order variational integrators in simulation and trajectory optimization of complex robotic systems. The efficacy of this paper is validated through comparison with existing methods, and implementation on various robotic systems—including trajectory optimization of the Spring Flamingo robot, the LittleDog robot and the Atlas robot. The results illustrate that the same integrator can be used for simulation and trajectory optimization in robotics, preserving mechanical properties while achieving good scalability and accuracy.
http://arxiv.org/abs/1904.12756
Segmenting clouds in high-resolution satellite images is an arduous and challenging task due to the many types of geographies and clouds a satellite can capture. Therefore, it needs to be automated and optimized, specially for those who regularly process great amounts of satellite images, such as governmental institutions. In that sense, the contribution of this work is twofold: We present the CloudPeru2 dataset, consisting of 22,400 images of 512x512 pixels and their respective hand-drawn cloud masks, as well as the proposal of an end-to-end segmentation method for clouds using a Convolutional Neural Network (CNN) based on the Deeplab v3+ architecture. The results over the test set achieved an accuracy of 96.62%, precision of 96.46%, specificity of 98.53%, and sensitivity of 96.72% which is superior to the compared methods.
http://arxiv.org/abs/1904.12743
In recent years, estimating the 6D pose of object instances with convolutional neural network (CNN) has received considerable attention. Depending on whether intermediate cues are used, the relevant literature can be roughly divided into two broad categories: direct methods and two stage pipelines. For the latter, intermediate cues, such as 3D object coordinates, semantic keypoints, or virtual control points instead of pose parameters are regressed by CNN in the first stage. Object pose can then be solved by correspondence constraints constructed with these intermediate cues. In this paper, we focus on the postprocessing of a two-stage pipeline and propose to combine two learning concepts for estimating object pose under challenging scenes: projection grouping on one side, and correspondence learning on the other. We firstly employ a local patch based method to predict projection heatmaps which denote the confidence distribution of projection of 3D bounding box’s corners. A projection grouping module is then proposed to remove redundant local maxima from each layer of heatmaps. Instead of directly feeding 2D-3D correspondences to the perspective-n-point (PnP) algorithm, multiple correspondence hypotheses are sampled from local maxima and its corresponding neighborhood and ranked by a correspondence-evaluation network. Finally, correspondences with higher confidence are selected to determine object pose. Extensive experiments on three public datasets demonstrate that the proposed framework outperforms several state of the art methods.
http://arxiv.org/abs/1904.12735
We propose a framework of genetic algorithms which use multi-level hierarchies to solve an optimization problem by searching over the space of simpler objective functions. We solve a variant of Travelling Salesman Problem called \texttt{soft-TSP} and show that when the constraints on the overall objective function are changed the algorithm adapts to churn out solutions for the changed objective. We use this idea to speed up learning by systematically altering the constraints to find a more globally optimal solution. We also use this framework to solve polynomial regression where the actual objective function is unknown but searching over space of available objective functions yields a good approximate solution.
http://arxiv.org/abs/1812.10308