October 17, 2019

3217 words 16 mins read

Paper Group ANR 861

The Lives of Bots. Optimizing the F-measure for Threshold-free Salient Object Detection. Global Semantic Consistency for Zero-Shot Learning. Adversarial Risk Bounds via Function Transformation. GeoSay: A Geometric Saliency for Extracting Buildings in Remote Sensing Images. An Integrated Soft Computing Approach to a Multi-biometric Security Model. I …

The Lives of Bots


Title	The Lives of Bots
Authors	R. Stuart Geiger
Abstract	Automated software agents — or bots — have long been an important part of how Wikipedia’s volunteer community of editors write, edit, update, monitor, and moderate content. In this paper, I discuss the complex social and technical environment in which Wikipedia’s bots operate. This paper focuses on the establishment and role of English Wikipedia’s bot policies and the Bot Approvals Group, a volunteer committee that reviews applications for new bots and helps resolve conflicts between Wikipedians about automation. In particular, I examine an early bot controversy over the first bot in Wikipedia to automatically enforce a social norm about how Wikipedian editors ought to interact in discussion spaces. As I show, bots enforce many rules in Wikipedia, but humans produce these bots and negotiate rules around their operation. Because of the openness of Wikipedia’s processes around automation, we can vividly observe the often-invisible human work involved in such algorithmic systems — in stark contrast to most other user-generated content platforms.
Tasks
Published	2018-10-22
URL	http://arxiv.org/abs/1810.09590v1
PDF	http://arxiv.org/pdf/1810.09590v1.pdf
PWC	https://paperswithcode.com/paper/the-lives-of-bots
Repo
Framework

Optimizing the F-measure for Threshold-free Salient Object Detection


Title	Optimizing the F-measure for Threshold-free Salient Object Detection
Authors	Kai Zhao, Shanghua Gao, Wenguan Wang, Ming-Ming Cheng
Abstract	Current CNN-based solutions to salient object detection (SOD) mainly rely on the optimization of cross-entropy loss (CELoss). Then the quality of detected saliency maps is often evaluated in terms of F-measure. In this paper, we investigate an interesting issue: can we consistently use the F-measure formulation in both training and evaluation for SOD? By reformulating the standard F-measure we propose the relaxed F-measure which is differentiable w.r.t the posterior and can be easily appended to the back of CNNs as the loss function. Compared to the conventional cross-entropy loss of which the gradients decrease dramatically in the saturated area, our loss function, named FLoss, holds considerable gradients even when the activation approaches the target. Consequently, the FLoss can continuously force the network to produce polarized activations. Comprehensive benchmarks on several popular datasets show that FLoss outperforms the state-of-the-art with a considerable margin. More specifically, due to the polarized predictions, our method is able to obtain high-quality saliency maps without carefully tuning the optimal threshold, showing significant advantages in real-world applications.
Tasks	Object Detection, Salient Object Detection
Published	2018-05-19
URL	https://arxiv.org/abs/1805.07567v2
PDF	https://arxiv.org/pdf/1805.07567v2.pdf
PWC	https://paperswithcode.com/paper/optimizing-the-f-measure-for-threshold-free
Repo
Framework

Global Semantic Consistency for Zero-Shot Learning


Title	Global Semantic Consistency for Zero-Shot Learning
Authors	Fan Wu, Kai Tian, Jihong Guan, Shuigeng Zhou
Abstract	In image recognition, there are many cases where training samples cannot cover all target classes. Zero-shot learning (ZSL) utilizes the class semantic information to classify samples of the unseen categories that have no corresponding samples contained in the training set. In this paper, we propose an end-to-end framework, called Global Semantic Consistency Network (GSC-Net for short), which makes complete use of the semantic information of both seen and unseen classes, to support effective zero-shot learning. We also adopt a soft label embedding loss to further exploit the semantic relationships among classes. To adapt GSC-Net to a more practical setting, Generalized Zero-shot Learning (GZSL), we introduce a parametric novelty detection mechanism. Our approach achieves the state-of-the-art performance on both ZSL and GZSL tasks over three visual attribute datasets, which validates the effectiveness and advantage of the proposed framework.
Tasks	Zero-Shot Learning
Published	2018-06-22
URL	http://arxiv.org/abs/1806.08503v1
PDF	http://arxiv.org/pdf/1806.08503v1.pdf
PWC	https://paperswithcode.com/paper/global-semantic-consistency-for-zero-shot
Repo
Framework

Adversarial Risk Bounds via Function Transformation


Title	Adversarial Risk Bounds via Function Transformation
Authors	Justin Khim, Po-Ling Loh
Abstract	We derive bounds for a notion of adversarial risk, designed to characterize the robustness of linear and neural network classifiers to adversarial perturbations. Specifically, we introduce a new class of function transformations with the property that the risk of the transformed functions upper-bounds the adversarial risk of the original functions. This reduces the problem of deriving bounds on the adversarial risk to the problem of deriving risk bounds using standard learning-theoretic techniques. We then derive bounds on the Rademacher complexities of the transformed function classes, obtaining error rates on the same order as the generalization error of the original function classes. We also discuss extensions of our theory to multiclass classification and regression. Finally, we provide two algorithms for optimizing the adversarial risk bounds in the linear case, and discuss connections to regularization and distributional robustness.
Tasks
Published	2018-10-22
URL	http://arxiv.org/abs/1810.09519v2
PDF	http://arxiv.org/pdf/1810.09519v2.pdf
PWC	https://paperswithcode.com/paper/adversarial-risk-bounds-via-function
Repo
Framework

GeoSay: A Geometric Saliency for Extracting Buildings in Remote Sensing Images


Title	GeoSay: A Geometric Saliency for Extracting Buildings in Remote Sensing Images
Authors	Gui-Song Xia, Jin Huang, Nan Xue, Qikai Lu, Xiaoxiang Zhu
Abstract	Automatic extraction of buildings in remote sensing images is an important but challenging task and finds many applications in different fields such as urban planning, navigation and so on. This paper addresses the problem of buildings extraction in very high-spatial-resolution (VHSR) remote sensing (RS) images, whose spatial resolution is often up to half meters and provides rich information about buildings. Based on the observation that buildings in VHSR-RS images are always more distinguishable in geometry than in texture or spectral domain, this paper proposes a geometric building index (GBI) for accurate building extraction, by computing the geometric saliency from VHSR-RS images. More precisely, given an image, the geometric saliency is derived from a mid-level geometric representations based on meaningful junctions that can locally describe geometrical structures of images. The resulting GBI is finally measured by integrating the derived geometric saliency of buildings. Experiments on three public and commonly used datasets demonstrate that the proposed GBI achieves the state-of-the-art performance and shows impressive generalization capability. Additionally, GBI preserves both the exact position and accurate shape of single buildings compared to existing methods.
Tasks	Extracting Buildings In Remote Sensing Images
Published	2018-11-07
URL	http://arxiv.org/abs/1811.02793v1
PDF	http://arxiv.org/pdf/1811.02793v1.pdf
PWC	https://paperswithcode.com/paper/geosay-a-geometric-saliency-for-extracting
Repo
Framework

An Integrated Soft Computing Approach to a Multi-biometric Security Model


Title	An Integrated Soft Computing Approach to a Multi-biometric Security Model
Authors	Prem Sewak Sudhish
Abstract	The abstract of the thesis consists of three sections, videlicet, Motivation Chapter Organization Salient Contributions. The complete abstract is included with the thesis. The final section on Salient Contributions is reproduced below. Salient Contributions The research presents the following salient contributions: i. A novel technique has been developed for comparing biographical information, by combining the average impact of Levenshtein, Damerau-Levenshtein, and editor distances. The impact is calculated as the ratio of the edit distance to the maximum possible edit distance between two strings of the same lengths as the given pair of strings. This impact lies in the range [0, 1] and can easily be converted to a similarity (matching) score by subtracting the impact from unity. ii. A universal soft computing framework is proposed for adaptively fusing biometric and biographical information by making real-time decisions to determine after consideration of each individual identifier whether computation of matching scores and subsequent fusion of additional identifiers, including biographical information is required. This proposed framework not only improves the accuracy of the system by fusing less reliable information (e.g. biographical information) only for instances where such a fusion is required, but also improves the efficiency of the system by computing matching scores for various available identifiers only when this computation is considered necessary. iii. A scientific method for comparing efficiency of fusion strategies through a predicted effort to error trade-off curve.
Tasks
Published	2018-01-25
URL	http://arxiv.org/abs/1801.08480v1
PDF	http://arxiv.org/pdf/1801.08480v1.pdf
PWC	https://paperswithcode.com/paper/an-integrated-soft-computing-approach-to-a
Repo
Framework

Improving Automated Latent Fingerprint Identification using Extended Minutia Types


Title	Improving Automated Latent Fingerprint Identification using Extended Minutia Types
Authors	Ram P. Krish, Julian Fierrez, Daniel Ramos, Fernando Alonso-Fernandez, Josef Bigun
Abstract	Latent fingerprints are usually processed with Automated Fingerprint Identification Systems (AFIS) by law enforcement agencies to narrow down possible suspects from a criminal database. AFIS do not commonly use all discriminatory features available in fingerprints but typically use only some types of features automatically extracted by a feature extraction algorithm. In this work, we explore ways to improve rank identification accuracies of AFIS when only a partial latent fingerprint is available. Towards solving this challenge, we propose a method that exploits extended fingerprint features (unusual/rare minutiae) not commonly considered in AFIS. This new method can be combined with any existing minutiae-based matcher. We first compute a similarity score based on least squares between latent and tenprint minutiae points, with rare minutiae features as reference points. Then the similarity score of the reference minutiae-based matcher at hand is modified based on a fitting error from the least square similarity stage. We use a realistic forensic fingerprint casework database in our experiments which contains rare minutiae features obtained from Guardia Civil, the Spanish law enforcement agency. Experiments are conducted using three minutiae-based matchers as a reference, namely: NIST-Bozorth3, VeriFinger-SDK and MCC-SDK. We report significant improvements in the rank identification accuracies when these minutiae matchers are augmented with our proposed algorithm based on rare minutiae features.
Tasks
Published	2018-10-23
URL	http://arxiv.org/abs/1810.09801v1
PDF	http://arxiv.org/pdf/1810.09801v1.pdf
PWC	https://paperswithcode.com/paper/improving-automated-latent-fingerprint
Repo
Framework

Robust Fuzzy-Learning For Partially Overlapping Channels Allocation In UAV Communication Networks


Title	Robust Fuzzy-Learning For Partially Overlapping Channels Allocation In UAV Communication Networks
Authors	Chaoqiong Fan, Bin Li, Jia Hou, Yi Wu, Weisi Guo, Chenglin Zhao
Abstract	In this paper, we consider a mesh-structured unmanned aerial vehicle (UAV) networks exploiting partially overlapping channels (POCs). For general data-collection tasks in UAV networks, we aim to optimize the network throughput with constraints on transmission power and quality of service (QoS). As far as the highly mobile and constantly changing UAV networks are concerned, unfortunately, most existing methods rely on definite information which is vulnerable to the dynamic environment, rendering system performance to be less effective. In order to combat dynamic topology and varying interference of UAV networks, a robust and distributed learning scheme is proposed. Rather than the perfect channel state information (CSI), we introduce uncertainties to characterize the dynamic channel gains among UAV nodes, which are then interpreted with fuzzy numbers. Instead of the traditional observation space where the channel capacity is a crisp reward, we implement the learning and decision process in a mapped fuzzy space. This allows the system to achieve a smoother and more robust performance by optimizing in an alternate space. To this end, we design a fuzzy payoffs function (FPF) to describe the fluctuated utility, and the problem of POCs assignment is formulated as a fuzzy payoffs game (FPG). Assisted by an attractive property of fuzzy bi-matrix games, the existence of fuzzy Nash equilibrium (FNE) for our formulated FPG is proved. Our robust fuzzy-learning algorithm could reach the equilibrium solution via a least-deviation method. Finally, numerical simulations are provided to demonstrate the advantages of our new scheme over the existing scheme.
Tasks
Published	2018-06-28
URL	http://arxiv.org/abs/1806.10756v1
PDF	http://arxiv.org/pdf/1806.10756v1.pdf
PWC	https://paperswithcode.com/paper/robust-fuzzy-learning-for-partially
Repo
Framework

Deep Learning Works in Practice. But Does it Work in Theory?


Title	Deep Learning Works in Practice. But Does it Work in Theory?
Authors	Lê Nguyên Hoang, Rachid Guerraoui
Abstract	Deep learning relies on a very specific kind of neural networks: those superposing several neural layers. In the last few years, deep learning achieved major breakthroughs in many tasks such as image analysis, speech recognition, natural language processing, and so on. Yet, there is no theoretical explanation of this success. In particular, it is not clear why the deeper the network, the better it actually performs. We argue that the explanation is intimately connected to a key feature of the data collected from our surrounding universe to feed the machine learning algorithms: large non-parallelizable logical depth. Roughly speaking, we conjecture that the shortest computational descriptions of the universe are algorithms with inherently large computation times, even when a large number of computers are available for parallelization. Interestingly, this conjecture, combined with the folklore conjecture in theoretical computer science that $ P \neq NC$, explains the success of deep learning.
Tasks	Speech Recognition
Published	2018-01-31
URL	http://arxiv.org/abs/1801.10437v1
PDF	http://arxiv.org/pdf/1801.10437v1.pdf
PWC	https://paperswithcode.com/paper/deep-learning-works-in-practice-but-does-it
Repo
Framework

Dynamic Regret Convergence Analysis and an Adaptive Regularization Algorithm for On-Policy Robot Imitation Learning


Title	Dynamic Regret Convergence Analysis and an Adaptive Regularization Algorithm for On-Policy Robot Imitation Learning
Authors	Jonathan N. Lee, Michael Laskey, Ajay Kumar Tanwani, Anil Aswani, Ken Goldberg
Abstract	On-policy imitation learning algorithms such as DAgger evolve a robot control policy by executing it, measuring performance (loss), obtaining corrective feedback from a supervisor, and generating the next policy. As the loss between iterations can vary unpredictably, a fundamental question is under what conditions this process will eventually achieve a converged policy. If one assumes the underlying trajectory distribution is static (stationary), it is possible to prove convergence for DAgger. However, in more realistic models for robotics, the underlying trajectory distribution is dynamic because it is a function of the policy. Recent results show it is possible to prove convergence of DAgger when a regularity condition on the rate of change of the trajectory distributions is satisfied. In this article, we reframe this result using dynamic regret theory from the field of online optimization and show that dynamic regret can be applied to any on-policy algorithm to analyze its convergence and optimality. These results inspire a new algorithm, Adaptive On-Policy Regularization (AOR), that ensures the conditions for convergence. We present simulation results with cart-pole balancing and locomotion benchmarks that suggest AOR can significantly decrease dynamic regret and chattering as the robot learns. To our knowledge, this the first application of dynamic regret theory to imitation learning.
Tasks	Imitation Learning
Published	2018-11-06
URL	https://arxiv.org/abs/1811.02184v2
PDF	https://arxiv.org/pdf/1811.02184v2.pdf
PWC	https://paperswithcode.com/paper/a-dynamic-regret-analysis-and-adaptive
Repo
Framework

Recovery Guarantees for Quadratic Tensors with Limited Observations


Title	Recovery Guarantees for Quadratic Tensors with Limited Observations
Authors	Hongyang Zhang, Vatsal Sharan, Moses Charikar, Yingyu Liang
Abstract	We consider the tensor completion problem of predicting the missing entries of a tensor. The commonly used CP model has a triple product form, but an alternate family of quadratic models which are the sum of pairwise products instead of a triple product have emerged from applications such as recommendation systems. Non-convex methods are the method of choice for learning quadratic models, and this work examines their sample complexity and error guarantee. Our main result is that with the number of samples being only linear in the dimension, all local minima of the mean squared error objective are global minima and recover the original tensor accurately. The techniques lead to simple proofs showing that convex relaxation can recover quadratic tensors provided with linear number of samples. We substantiate our theoretical results with experiments on synthetic and real-world data, showing that quadratic models have better performance than CP models in scenarios where there are limited amount of observations available.
Tasks	Recommendation Systems
Published	2018-10-31
URL	http://arxiv.org/abs/1811.00148v1
PDF	http://arxiv.org/pdf/1811.00148v1.pdf
PWC	https://paperswithcode.com/paper/recovery-guarantees-for-quadratic-tensors
Repo
Framework

Comparing Graph Clusterings: Set partition measures vs. Graph-aware measures


Title	Comparing Graph Clusterings: Set partition measures vs. Graph-aware measures
Authors	Valérie Poulin, François Théberge
Abstract	In this paper, we propose a family of graph partition similarity measures that take the topology of the graph into account. These graph-aware measures are alternatives to using set partition similarity measures that are not specifically designed for graph partitions. The two types of measures, graph-aware and set partition measures, are shown to have opposite behaviors with respect to resolution issues and provide complementary information necessary to assess that two graph partitions are similar.
Tasks
Published	2018-06-29
URL	http://arxiv.org/abs/1806.11494v2
PDF	http://arxiv.org/pdf/1806.11494v2.pdf
PWC	https://paperswithcode.com/paper/comparing-graph-clusterings-set-partition
Repo
Framework

Domain Adaptation for Robot Predictive Maintenance Systems


Title	Domain Adaptation for Robot Predictive Maintenance Systems
Authors	Arash Golibagh Mahyari, Thomas Locker
Abstract	Industrial robots play an increasingly important role in a growing number of fields. For example, robotics is used to increase productivity while reducing costs in various aspects of manufacturing. Since robots are often set up in production lines, the breakdown of a single robot has a negative impact on the entire process, in the worst case bringing the whole line to a halt until the issue is resolved, leading to substantial financial losses due to the unforeseen downtime. Therefore, predictive maintenance systems based on the internal signals of robots have gained attention as an essential component of robotics service offerings. The main shortcoming of existing predictive maintenance algorithms is that the extracted features typically differ significantly from the learnt model when the operation of the robot changes, incurring false alarms. In order to mitigate this problem, predictive maintenance algorithms require the model to be retrained with normal data of the new operation. In this paper, we propose a novel solution based on transfer learning to pass the knowledge of the trained model from one operation to another in order to prevent the need for retraining and to eliminate such false alarms. The deployment of the proposed unsupervised transfer learning algorithm on real-world datasets demonstrates that the algorithm can not only distinguish between operation and mechanical condition change, it further yields a sharper deviation from the trained model in case of a mechanical condition change and thus detects mechanical issues with higher confidence.
Tasks	Domain Adaptation, Transfer Learning
Published	2018-09-23
URL	https://arxiv.org/abs/1809.08626v3
PDF	https://arxiv.org/pdf/1809.08626v3.pdf
PWC	https://paperswithcode.com/paper/domain-adaptation-in-robot-fault-diagnostic
Repo
Framework


Title	Recipe1M+: A Dataset for Learning Cross-Modal Embeddings for Cooking Recipes and Food Images
Authors	Javier Marin, Aritro Biswas, Ferda Ofli, Nicholas Hynes, Amaia Salvador, Yusuf Aytar, Ingmar Weber, Antonio Torralba
Abstract	In this paper, we introduce Recipe1M+, a new large-scale, structured corpus of over one million cooking recipes and 13 million food images. As the largest publicly available collection of recipe data, Recipe1M+ affords the ability to train high-capacity modelson aligned, multimodal data. Using these data, we train a neural network to learn a joint embedding of recipes and images that yields impressive results on an image-recipe retrieval task. Moreover, we demonstrate that regularization via the addition of a high-level classification objective both improves retrieval performance to rival that of humans and enables semantic vector arithmetic. We postulate that these embeddings will provide a basis for further exploration of the Recipe1M+ dataset and food and cooking in general. Code, data and models are publicly available.
Tasks
Published	2018-10-14
URL	https://arxiv.org/abs/1810.06553v2
PDF	https://arxiv.org/pdf/1810.06553v2.pdf
PWC	https://paperswithcode.com/paper/recipe1m-a-dataset-for-learning-cross-modal
Repo
Framework

OntoWind: An Improved and Extended Wind Energy Ontology


Title	OntoWind: An Improved and Extended Wind Energy Ontology
Authors	Dilek Küçük, Doğan Küçük
Abstract	Ontologies are critical sources of semantic information for many application domains. Hence, there are ontologies proposed and utilized for domains such as medicine, chemical engineering, and electrical energy. In this paper, we present an improved and extended version of a wind energy ontology previously proposed. First, the ontology is restructured to increase its understandability and coverage. Secondly, it is enriched with new concepts, crisp/fuzzy attributes, and instances to increase its usability in semantic applications regarding wind energy. The ultimate ontology is utilized within a Web-based semantic portal application for wind energy, in order to showcase its contribution in a genuine application. Hence, the current study is a significant to wind and thereby renewable energy informatics, with the presented publicly-available wind energy ontology and the implemented proof-of-concept system.
Tasks
Published	2018-03-07
URL	http://arxiv.org/abs/1803.02808v1
PDF	http://arxiv.org/pdf/1803.02808v1.pdf
PWC	https://paperswithcode.com/paper/ontowind-an-improved-and-extended-wind-energy
Repo
Framework