October 19, 2019

2752 words 13 mins read

Paper Group ANR 159

Generalized Scene Reconstruction. Taskonomy: Disentangling Task Transfer Learning. A Dynamic Model for Traffic Flow Prediction Using Improved DRN. SA-IGA: A Multiagent Reinforcement Learning Method Towards Socially Optimal Outcomes. Cardinality Leap for Open-Ended Evolution: Theoretical Consideration and Demonstration by “Hash Chemistry”. GAPLE: Ge …

Generalized Scene Reconstruction


Title	Generalized Scene Reconstruction
Authors	John K. Leffingwell, Donald J. Meagher, Khan W. Mahmud, Scott Ackerson
Abstract	A new passive approach called Generalized Scene Reconstruction (GSR) enables “generalized scenes” to be effectively reconstructed. Generalized scenes are defined to be “boundless” spaces that include non-Lambertian, partially transmissive, textureless and finely-structured matter. A new data structure called a plenoptic octree is introduced to enable efficient (database-like) light and matter field reconstruction in devices such as mobile phones, augmented reality (AR) glasses and drones. To satisfy threshold requirements for GSR accuracy, scenes are represented as systems of partially polarized light, radiometrically interacting with matter. To demonstrate GSR, a prototype imaging polarimeter is used to reconstruct (in generalized light fields) highly reflective, hail-damaged automobile body panels. Follow-on GSR experiments are described.
Tasks
Published	2018-03-22
URL	http://arxiv.org/abs/1803.08496v3
PDF	http://arxiv.org/pdf/1803.08496v3.pdf
PWC	https://paperswithcode.com/paper/generalized-scene-reconstruction
Repo
Framework

Taskonomy: Disentangling Task Transfer Learning


Title	Taskonomy: Disentangling Task Transfer Learning
Authors	Amir Zamir, Alexander Sax, William Shen, Leonidas Guibas, Jitendra Malik, Silvio Savarese
Abstract	Do visual tasks have a relationship, or are they unrelated? For instance, could having surface normals simplify estimating the depth of an image? Intuition answers these questions positively, implying existence of a structure among visual tasks. Knowing this structure has notable values; it is the concept underlying transfer learning and provides a principled way for identifying redundancies across tasks, e.g., to seamlessly reuse supervision among related tasks or solve many tasks in one system without piling up the complexity. We proposes a fully computational approach for modeling the structure of space of visual tasks. This is done via finding (first and higher-order) transfer learning dependencies across a dictionary of twenty six 2D, 2.5D, 3D, and semantic tasks in a latent space. The product is a computational taxonomic map for task transfer learning. We study the consequences of this structure, e.g. nontrivial emerged relationships, and exploit them to reduce the demand for labeled data. For example, we show that the total number of labeled datapoints needed for solving a set of 10 tasks can be reduced by roughly 2/3 (compared to training independently) while keeping the performance nearly the same. We provide a set of tools for computing and probing this taxonomical structure including a solver that users can employ to devise efficient supervision policies for their use cases.
Tasks	Transfer Learning
Published	2018-04-23
URL	http://arxiv.org/abs/1804.08328v1
PDF	http://arxiv.org/pdf/1804.08328v1.pdf
PWC	https://paperswithcode.com/paper/taskonomy-disentangling-task-transfer
Repo
Framework

A Dynamic Model for Traffic Flow Prediction Using Improved DRN


Title	A Dynamic Model for Traffic Flow Prediction Using Improved DRN
Authors	Zeren Tan, Ruimin Li
Abstract	Real-time traffic flow prediction can not only provide travelers with reliable traffic information so that it can save people’s time, but also assist the traffic management agency to manage traffic system. It can greatly improve the efficiency of the transportation system. Traditional traffic flow prediction approaches usually need a large amount of data but still give poor performances. With the development of deep learning, researchers begin to pay attention to artificial neural networks (ANNs) such as RNN and LSTM. However, these ANNs are very time-consuming. In our research, we improve the Deep Residual Network and build a dynamic model which previous researchers hardly use. We firstly integrate the input and output of the $i^{th}$ layer to the input of the $i+1^{th}$ layer and prove that each layer will fit a simpler function so that the error rate will be much smaller. Then, we use the concept of online learning in our model to update pre-trained model during prediction. Our result shows that our model has higher accuracy than some state-of-the-art models. In addition, our dynamic model can perform better in practical applications.
Tasks
Published	2018-05-02
URL	http://arxiv.org/abs/1805.00868v4
PDF	http://arxiv.org/pdf/1805.00868v4.pdf
PWC	https://paperswithcode.com/paper/a-dynamic-model-for-traffic-flow-prediction
Repo
Framework

SA-IGA: A Multiagent Reinforcement Learning Method Towards Socially Optimal Outcomes


Title	SA-IGA: A Multiagent Reinforcement Learning Method Towards Socially Optimal Outcomes
Authors	Chengwei Zhang, Xiaohong Li, Jianye Hao, Siqi Chen, Karl Tuyls, Wanli Xue
Abstract	In multiagent environments, the capability of learning is important for an agent to behave appropriately in face of unknown opponents and dynamic environment. From the system designer’s perspective, it is desirable if the agents can learn to coordinate towards socially optimal outcomes, while also avoiding being exploited by selfish opponents. To this end, we propose a novel gradient ascent based algorithm (SA-IGA) which augments the basic gradient-ascent algorithm by incorporating social awareness into the policy update process. We theoretically analyze the learning dynamics of SA-IGA using dynamical system theory and SA-IGA is shown to have linear dynamics for a wide range of games including symmetric games. The learning dynamics of two representative games (the prisoner’s dilemma game and the coordination game) are analyzed in details. Based on the idea of SA-IGA, we further propose a practical multiagent learning algorithm, called SA-PGA, based on Q-learning update rule. Simulation results show that SA-PGA agent can achieve higher social welfare than previous social-optimality oriented Conditional Joint Action Learner (CJAL) and also is robust against individually rational opponents by reaching Nash equilibrium solutions.
Tasks	Q-Learning
Published	2018-03-08
URL	http://arxiv.org/abs/1803.03021v1
PDF	http://arxiv.org/pdf/1803.03021v1.pdf
PWC	https://paperswithcode.com/paper/sa-iga-a-multiagent-reinforcement-learning
Repo
Framework

Cardinality Leap for Open-Ended Evolution: Theoretical Consideration and Demonstration by “Hash Chemistry”


Title	Cardinality Leap for Open-Ended Evolution: Theoretical Consideration and Demonstration by “Hash Chemistry”
Authors	Hiroki Sayama
Abstract	Open-ended evolution requires unbounded possibilities that evolving entities can explore. The cardinality of a set of those possibilities thus has a significant implication for the open-endedness of evolution. We propose that facilitating formation of higher-order entities is a generalizable, effective way to cause a “cardinality leap” in the set of possibilities that promotes open-endedness. We demonstrate this idea with a simple, proof-of-concept toy model called “Hash Chemistry” that uses a hash function as a fitness evaluator of evolving entities of any size/order. Simulation results showed that the cumulative number of unique replicating entities that appeared in evolution increased almost linearly along time without an apparent bound, demonstrating the effectiveness of the proposed cardinality leap. It was also observed that the number of individual entities involved in a single replication event gradually increased over time, indicating evolutionary appearance of higher-order entities. Moreover, these behaviors were not observed in control experiments in which fitness evaluators were replaced by random number generators. This strongly suggests that the dynamics observed in Hash Chemistry were indeed evolutionary behaviors driven by selection and adaptation taking place at multiple scales.
Tasks
Published	2018-06-18
URL	http://arxiv.org/abs/1806.06628v4
PDF	http://arxiv.org/pdf/1806.06628v4.pdf
PWC	https://paperswithcode.com/paper/cardinality-leap-for-open-ended-evolution
Repo
Framework

GAPLE: Generalizable Approaching Policy LEarning for Robotic Object Searching in Indoor Environment


Title	GAPLE: Generalizable Approaching Policy LEarning for Robotic Object Searching in Indoor Environment
Authors	Xin Ye, Zhe Lin, Joon-Young Lee, Jianming Zhang, Shibin Zheng, Yezhou Yang
Abstract	We study the problem of learning a generalizable action policy for an intelligent agent to actively approach an object of interest in an indoor environment solely from its visual inputs. While scene-driven or recognition-driven visual navigation has been widely studied, prior efforts suffer severely from the limited generalization capability. In this paper, we first argue the object searching task is environment dependent while the approaching ability is general. To learn a generalizable approaching policy, we present a novel solution dubbed as GAPLE which adopts two channels of visual features: depth and semantic segmentation, as the inputs to the policy learning module. The empirical studies conducted on the House3D dataset as well as on a physical platform in a real world scenario validate our hypothesis, and we further provide in-depth qualitative analysis.
Tasks	Semantic Segmentation, Visual Navigation
Published	2018-09-21
URL	http://arxiv.org/abs/1809.08287v2
PDF	http://arxiv.org/pdf/1809.08287v2.pdf
PWC	https://paperswithcode.com/paper/gaple-generalizable-approaching-policy
Repo
Framework

idtracker.ai: Tracking all individuals in large collectives of unmarked animals


Title	idtracker.ai: Tracking all individuals in large collectives of unmarked animals
Authors	Francisco Romero-Ferrero, Mattia G. Bergomi, Robert Hinz, Francisco J. H. Heras, Gonzalo G. de Polavieja
Abstract	Our understanding of collective animal behavior is limited by our ability to track each of the individuals. We describe an algorithm and software, idtracker.ai, that extracts from video all trajectories with correct identities at a high accuracy for collectives of up to 100 individuals. It uses two deep networks, one detecting when animals touch or cross and another one for animal identification, trained adaptively to conditions and difficulty of the video.
Tasks
Published	2018-03-12
URL	http://arxiv.org/abs/1803.04351v1
PDF	http://arxiv.org/pdf/1803.04351v1.pdf
PWC	https://paperswithcode.com/paper/idtrackerai-tracking-all-individuals-in-large
Repo
Framework

A Deeper Look at 3D Shape Classifiers


Title	A Deeper Look at 3D Shape Classifiers
Authors	Jong-Chyi Su, Matheus Gadelha, Rui Wang, Subhransu Maji
Abstract	We investigate the role of representations and architectures for classifying 3D shapes in terms of their computational efficiency, generalization, and robustness to adversarial transformations. By varying the number of training examples and employing cross-modal transfer learning we study the role of initialization of existing deep architectures for 3D shape classification. Our analysis shows that multiview methods continue to offer the best generalization even without pretraining on large labeled image datasets, and even when trained on simplified inputs such as binary silhouettes. Furthermore, the performance of voxel-based 3D convolutional networks and point-based architectures can be improved via cross-modal transfer from image representations. Finally, we analyze the robustness of 3D shape classifiers to adversarial transformations and present a novel approach for generating adversarial perturbations of a 3D shape for multiview classifiers using a differentiable renderer. We find that point-based networks are more robust to point position perturbations while voxel-based and multiview networks are easily fooled with the addition of imperceptible noise to the input.
Tasks	Transfer Learning
Published	2018-09-07
URL	http://arxiv.org/abs/1809.02560v2
PDF	http://arxiv.org/pdf/1809.02560v2.pdf
PWC	https://paperswithcode.com/paper/a-deeper-look-at-3d-shape-classifiers
Repo
Framework

Asymptotic performance of regularized multi-task learning


Title	Asymptotic performance of regularized multi-task learning
Authors	Shaohan Chen, Chuanhou Gao
Abstract	This paper analyzes asymptotic performance of a regularized multi-task learning model where task parameters are optimized jointly. If tasks are closely related, empirical work suggests multi-task learning models to outperform single-task ones in finite sample cases. As data size grows indefinitely, we show the learned multi-classifier to optimize an average misclassification error function which depicts the risk of applying multi-task learning algorithm to making decisions. This technique conclusion demonstrates the regularized multi-task learning model to be able to produce reliable decision rule for each task in the sense that it will asymptotically converge to the corresponding Bayes rule. Also, we find the interaction effect between tasks vanishes as data size growing indefinitely, which is quite different from the behavior in finite sample cases.
Tasks	Multi-Task Learning
Published	2018-05-31
URL	http://arxiv.org/abs/1805.12507v1
PDF	http://arxiv.org/pdf/1805.12507v1.pdf
PWC	https://paperswithcode.com/paper/asymptotic-performance-of-regularized-multi
Repo
Framework

Augmented Space Linear Model


Title	Augmented Space Linear Model
Authors	Zhengda Qin, Badong Chen, Nanning Zheng, Jose C. Principe
Abstract	The linear model uses the space defined by the input to project the target or desired signal and find the optimal set of model parameters. When the problem is nonlinear, the adaption requires nonlinear models for good performance, but it becomes slower and more cumbersome. In this paper, we propose a linear model called Augmented Space Linear Model (ASLM), which uses the full joint space of input and desired signal as the projection space and approaches the performance of nonlinear models. This new algorithm takes advantage of the linear solution, and corrects the estimate for the current testing phase input with the error assigned to the input space neighborhood in the training phase. This algorithm can solve the nonlinear problem with the computational efficiency of linear methods, which can be regarded as a trade off between accuracy and computational complexity. Making full use of the training data, the proposed augmented space model may provide a new way to improve many modeling tasks.
Tasks
Published	2018-02-01
URL	http://arxiv.org/abs/1802.00174v2
PDF	http://arxiv.org/pdf/1802.00174v2.pdf
PWC	https://paperswithcode.com/paper/augmented-space-linear-model
Repo
Framework

Faithful Multimodal Explanation for Visual Question Answering


Title	Faithful Multimodal Explanation for Visual Question Answering
Authors	Jialin Wu, Raymond J. Mooney
Abstract	AI systems’ ability to explain their reasoning is critical to their utility and trustworthiness. Deep neural networks have enabled significant progress on many challenging problems such as visual question answering (VQA). However, most of them are opaque black boxes with limited explanatory capability. This paper presents a novel approach to developing a high-performing VQA system that can elucidate its answers with integrated textual and visual explanations that faithfully reflect important aspects of its underlying reasoning while capturing the style of comprehensible human explanations. Extensive experimental evaluation demonstrates the advantages of this approach compared to competing methods with both automatic evaluation metrics and human evaluation metrics.
Tasks	Question Answering, Visual Question Answering
Published	2018-09-08
URL	https://arxiv.org/abs/1809.02805v2
PDF	https://arxiv.org/pdf/1809.02805v2.pdf
PWC	https://paperswithcode.com/paper/faithful-multimodal-explanation-for-visual
Repo
Framework

SafeDrive: Enhancing Lane Appearance for Autonomous and Assisted Driving Under Limited Visibility


Title	SafeDrive: Enhancing Lane Appearance for Autonomous and Assisted Driving Under Limited Visibility
Authors	Jiawei Mo, Junaed Sattar
Abstract	Autonomous detection of lane markers improves road safety, and purely visual tracking is desirable for widespread vehicle compatibility and reducing sensor intrusion, cost, and energy consumption. However, visual approaches are often ineffective because of a number of factors; e.g., occlusion, poor weather conditions, and paint wear-off. We present an approach to enhance lane marker appearance for assisted and autonomous driving, particularly under poor visibility. Our method, named SafeDrive, attempts to improve visual lane detection approaches in drastically degraded visual conditions. SafeDrive finds lane markers in alternate imagery of the road at the vehicle’s location and reconstructs a sparse 3D model of the surroundings. By estimating the geometric relationship between this 3D model and the current view, the lane markers are projected onto the visual scene; any lane detection algorithm can be subsequently used to detect lanes in the resulting image. SafeDrive does not require additional sensors other than vision and location data. We demonstrate the effectiveness of our approach on a number of test cases obtained from actual driving data recorded in urban settings.
Tasks	Autonomous Driving, Lane Detection, Visual Tracking
Published	2018-07-24
URL	http://arxiv.org/abs/1807.11575v1
PDF	http://arxiv.org/pdf/1807.11575v1.pdf
PWC	https://paperswithcode.com/paper/safedrive-enhancing-lane-appearance-for
Repo
Framework

FineTag: Multi-attribute Classification at Fine-grained Level in Images


Title	FineTag: Multi-attribute Classification at Fine-grained Level in Images
Authors	Roshanak Zakizadeh, Michele Sasdelli, Yu Qian, Eduard Vazquez
Abstract	In this paper, we address the extraction of the fine-grained attributes of an instance as a `multi-attribute classification’ problem. To this end, we propose an end-to-end architecture by adopting the bi-linear Convolutional Neural Network with the pairwise ranking loss. This is the first time such architecture is applied for the fine-grained attributes classification problem. We compared the proposed method with a competitive deep Convolutional Neural Network baseline. Extensive experiments show that the proposed method attains/outperforms the performance of compared baseline with significantly less number of parameters ($40\times$ less). We demonstrated our approach on CUB200 birds dataset whose annotations are adapted in this work for multi-attribute classification at fine-grained level. \|
Tasks
Published	2018-06-19
URL	http://arxiv.org/abs/1806.07124v2
PDF	http://arxiv.org/pdf/1806.07124v2.pdf
PWC	https://paperswithcode.com/paper/finetag-multi-attribute-classification-at
Repo
Framework

Adversarial Sampling for Active Learning


Title	Adversarial Sampling for Active Learning
Authors	Christoph Mayer, Radu Timofte
Abstract	This paper proposes asal, a new GAN based active learning method that generates high entropy samples. Instead of directly annotating the synthetic samples, ASAL searches similar samples from the pool and includes them for training. Hence, the quality of new samples is high and annotations are reliable. To the best of our knowledge, ASAL is the first GAN based AL method applicable to multi-class problems that outperforms random sample selection. Another benefit of ASAL is its small run-time complexity (sub-linear) compared to traditional uncertainty sampling (linear). We present a comprehensive set of experiments on multiple traditional data sets and show that ASAL outperforms similar methods and clearly exceeds the established baseline (random sampling). In the discussion section we analyze in which situations ASAL performs best and why it is sometimes hard to outperform random sample selection.
Tasks	Active Learning, Multi-Label Classification
Published	2018-08-20
URL	https://arxiv.org/abs/1808.06671v2
PDF	https://arxiv.org/pdf/1808.06671v2.pdf
PWC	https://paperswithcode.com/paper/adversarial-sampling-for-active-learning
Repo
Framework

Reversible Adversarial Examples


Title	Reversible Adversarial Examples
Authors	Jiayang Liu, Dongdong Hou, Weiming Zhang, Nenghai Yu
Abstract	Deep Neural Networks have recently led to significant improvement in many fields such as image classification and speech recognition. However, these machine learning models are vulnerable to adversarial examples which can mislead machine learning classifiers to give incorrect classifications. In this paper, we take advantage of reversible data hiding to construct reversible adversarial examples which are still misclassified by Deep Neural Networks. Furthermore, the proposed method can recover original images from reversible adversarial examples with no distortion.
Tasks	Image Classification, Speech Recognition
Published	2018-11-01
URL	http://arxiv.org/abs/1811.00189v2
PDF	http://arxiv.org/pdf/1811.00189v2.pdf
PWC	https://paperswithcode.com/paper/reversible-adversarial-examples
Repo
Framework