PyTorchKR Papers

최신 논문

총 3,337편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2106.10483

Cooperative Magneto-Inductive Localization

Wireless localization is a key requirement for many applications. It concerns position estimation of mobile nodes (agents) relative to fixed nodes (anchors) from wireless channel measurements. Cooperative localization is an advanced concept that considers the joint estimation of multiple agent positions based on channel measurements of all agent-anchor links together with all agent-agent links. In this paper we present the first study of cooperative localization for magneto-inductive wireless sensor networks, which are of technological interest due to good material penetration and channel predictability. We demonstrate significant accuracy improvements (a factor of 3 for 10 cooperating agents) over the non-cooperative scheme. The evaluation uses the Cram\'er-Rao lower bound on the cooperative position estimation error, which is derived herein. To realize this accuracy, the maximum-likelihood estimate (MLE) must be computed by solving a high-dimensional least-squares problem, whereby convergence to local minima proves to be problematic. A proposed cooperative localization algorithm addresses this issue: first, preliminary estimates of the agent positions and orientations are computed, which then serve as initial values for a gradient search. In all our test cases, this method yields the MLE and the associated high accuracy (comprising the cooperation gain) from a single solver run. The preliminary estimates use novel closed-form MLE formulas of the distance, direction and orientation for single links between three-axis coils, which are given in detail.

arXiv:2106.10766

Learning to Track Object Position through Occlusion

Occlusion is one of the most significant challenges encountered by object detectors and trackers. While both object detection and tracking has received a lot of attention in the past, most existing methods in this domain do not target detecting or tracking objects when they are occluded. However, being able to detect or track an object of interest through occlusion has been a long standing challenge for different autonomous tasks. Traditional methods that employ visual object trackers with explicit occlusion modeling experience drift and make several fundamental assumptions about the data. We propose to address this with a `tracking-by-detection` approach that builds upon the success of region based video object detectors. Our video level object detector uses a novel recurrent computational unit at its core that enables long term propagation of object features even under occlusion. Finally, we compare our approach with existing state-of-the-art video object detectors and show that our approach achieves superior results on a dataset of furniture assembly videos collected from the internet, where small objects like screws, nuts, and bolts often get occluded from the camera viewpoint.

arXiv:2106.09927

A symmetry-inclusive algebraic approach to genome rearrangement

Of the many modern approaches to calculating evolutionary distance via models of genome rearrangement, most are tied to a particular set of genomic modelling assumptions and to a restricted class of allowed rearrangements. The "position paradigm", in which genomes are represented as permutations signifying the position (and orientation) of each region, enables a refined model-based approach, where one can select biologically plausible rearrangements and assign to them relative probabilities/costs. Here, one must further incorporate any underlying structural symmetry of the genomes into the calculations and ensure that this symmetry is reflected in the model. In our recently-introduced framework of {\em genome algebras}, each genome corresponds to an element that simultaneously incorporates all of its inherent physical symmetries. The representation theory of these algebras then provides a natural model of evolution via rearrangement as a Markov chain. Whilst the implementation of this framework to calculate distances for genomes with `practical' numbers of regions is currently computationally infeasible, we consider it to be a significant theoretical advance: one can incorporate different genomic modelling assumptions, calculate various genomic distances, and compare the results under different rearrangement models. The aim of this paper is to demonstrate some of these features.

arXiv:2106.09616

Deep Reinforcement Learning Based Optimization for IRS Based UAV-NOMA Downlink Networks

This paper investigates the application of deep deterministic policy gradient (DDPG) to intelligent reflecting surface (IRS) based unmanned aerial vehicles (UAV) assisted non-orthogonal multiple access (NOMA) downlink networks. The deployment of the UAV equipped with an IRS is important, as the UAV increases the flexibility of the IRS significantly, especially for the case of users who have no line of sight (LoS) path to the base station (BS). Therefore, the aim of this letter is to maximize the sum rate by jointly optimizing the power allocation of the BS, the phase shifting of the IRS and the horizontal position of the UAV. Because the formulated problem is not convex, the DDPG algorithm is utilized to solve it. The computer simulation results are provided to show the superior performance of the proposed DDPG based algorithm.

arXiv:2106.09252

Temporal Logic Planning for Minimum-Time Positioning of Multiple Threat-Seduction Decoys

Reusable decoys offer a cost-effective alternative to the single-use hardware commonly applied to protect surface assets from threats. Such decoys portray fake assets to lure threats away from the true asset. To deceive a threat, a decoy first has to position itself such that it can break the radar lock. Considering multiple simultaneous threats, this paper introduces an approach for controlling multiple decoys to minimise the time required to break the locks of all the threats. The method includes the optimal allocation of one decoy to every threat with an assignment procedure that provides local position constraints to guarantee collision avoidance and thereby decouples the control of the decoys. A crude model of a decoy with uncertainty is considered for motion planning. The task of a decoy reaching a state in which the lock of the assigned threat can be broken is formulated as a temporal logic specification. To this end, the requirements to complete the task are modelled as time-varying set-membership constraints. The temporal and logical combination of the constraints is encoded in a mixed-integer optimisation problem. To demonstrate the results a simulated case study is provided.

arXiv:2106.09556

A Deep Reinforcement Learning Approach towards Pendulum Swing-up Problem based on TF-Agents

Adapting the idea of training CartPole with Deep Q-learning agent, we are able to find a promising result that prevent the pole from falling down. The capacity of reinforcement learning (RL) to learn from the interaction between the environment and agent provides an optimal control strategy. In this paper, we aim to solve the classic pendulum swing-up problem that making the learned pendulum to be in upright position and balanced. Deep Deterministic Policy Gradient algorithm is introduced to operate over continuous action domain in this problem. Salient results of optimal pendulum are proved with increasing average return, decreasing loss, and live video in the code part.

arXiv:2106.09258

Knowledge Graphs and Machine Learning in biased C4I applications

This paper introduces our position on the critical issue of bias that recently appeared in AI applications. Specifically, we discuss the combination of current technologies used in AI applications i.e., Machine Learning and Knowledge Graphs, and point to their involvement in (de)biased applications of the C4I domain. Although this is a wider problem that currently emerges from different application domains, bias appears more critical in C4I than in others due to its security-related nature. While proposing certain actions to be taken towards debiasing C4I applications, we acknowledge the immature aspect of this topic within the Knowledge Graph and Semantic Web communities.

arXiv:2106.08634

MPC-based Reinforcement Learning for a Simplified Freight Mission of Autonomous Surface Vehicles

In this work, we propose a Model Predictive Control (MPC)-based Reinforcement Learning (RL) method for Autonomous Surface Vehicles (ASVs). The objective is to find an optimal policy that minimizes the closed-loop performance of a simplified freight mission, including collision-free path following, autonomous docking, and a skillful transition between them. We use a parametrized MPC-scheme to approximate the optimal policy, which considers path-following/docking costs and states (position, velocity)/inputs (thruster force, angle) constraints. The Least Squares Temporal Difference (LSTD)-based Deterministic Policy Gradient (DPG) method is then applied to update the policy parameters. Our simulation results demonstrate that the proposed MPC-LSTD-based DPG method could improve the closed-loop performance during learning for the freight mission problem of ASV.

arXiv:2106.08609

Reinforcement learning for pursuit and evasion of microswimmers at low Reynolds number

We consider a model of two competing microswimming agents engaged in a pursue-evasion task within a low-Reynolds-number environment. Agents can only perform simple maneuvers and sense hydrodynamic disturbances, which provide ambiguous (partial) information about the opponent's position and motion. We frame the problem as a zero-sum game: The pursuer has to capture the evader in the shortest time, while the evader aims at deferring capture as long as possible. We show that the agents, trained via adversarial reinforcement learning, are able to overcome partial observability by discovering increasingly complex sequences of moves and countermoves that outperform known heuristic strategies and exploit the hydrodynamic environment.

arXiv:2106.08009공식 코드 · Papers with Code

Compositional Sketch Search

We present an algorithm for searching image collections using free-hand sketches that describe the appearance and relative positions of multiple objects. Sketch based image retrieval (SBIR) methods predominantly match queries containing a single, dominant object invariant to its position within an image. Our work exploits drawings as a concise and intuitive representation for specifying entire scene compositions. We train a convolutional neural network (CNN) to encode masked visual features from sketched objects, pooling these into a spatial descriptor encoding the spatial relationships and appearances of objects in the composition. Training the CNN backbone as a Siamese network under triplet loss yields a metric search embedding for measuring compositional similarity which may be efficiently leveraged for visual search by applying product quantization.

arXiv:2106.06258

DebiasGAN: Eliminating Position Bias in News Recommendation with Adversarial Learning

News recommendation is important for improving news reading experience of users. Users' news click behaviors are widely used for inferring user interests and predicting future clicks. However, click behaviors are heavily affected by the biases brought by the positions of news displayed on the webpage. It is important to eliminate the effect of position biases on the recommendation model to accurately target user interests. In this paper, we propose a news recommendation method named DebiasGAN that can effectively eliminate the effect of position biases via adversarial learning. We use a bias-aware click model to capture the influence of position bias on click behaviors, and we use a bias-invariant click model with random candidate news positions to estimate the ideally unbiased click scores. We apply adversarial learning techniques to the hidden representations learned by the two models to help the bias-invariant click model capture the bias-independent interest of users on news. Experimental results on two real-world datasets show that DebiasGAN can effectively improve the accuracy of news recommendation by eliminating position biases.

arXiv:2106.05716

Fastening the Initial Access in 5G NR Sidelink for 6G V2X Networks

The ever-increasing demand for intelligent, automated, and connected mobility solutions pushes for the development of an innovative sixth Generation (6G) of cellular networks. A radical transformation on the physical layer of vehicular communications is planned, with a paradigm shift towards beam-based millimeter Waves or sub-Terahertz communications, which require precise beam pointing for guaranteeing the communication link, especially in high mobility. A key design aspect is a fast and proactive Initial Access (IA) algorithm to select the optimal beam to be used. In this work, we investigate alternative IA techniques to fasten the current fifth-generation (5G) standard, targeting an efficient 6G design. First, we discuss cooperative position-based schemes that rely on the position information. Then, motivated by the intuition of a non-uniform distribution of the communication directions due to road topology constraints, we design two Probabilistic Codebook (PCB) techniques of prioritized beams. In the first one, the PCBs are built leveraging past collected traffic information, while in the second one, we use the Hough Transform over the digital map to extract dominant road directions. We also show that the information coming from the angular probability distribution allows designing non-uniform codebook quantization, reducing the degradation of the performances compared to uniform one. Numerical simulation on realistic scenarios shows that PCBs-based beam selection outperforms the 5G standard in terms of the number of IA trials, with a performance comparable to position-based methods, without requiring the signaling of sensitive information.

arXiv:2106.05505공식 코드 · Papers with Code

Convolutions and Self-Attention: Re-interpreting Relative Positions in Pre-trained Language Models

In this paper, we detail the relationship between convolutions and self-attention in natural language tasks. We show that relative position embeddings in self-attention layers are equivalent to recently-proposed dynamic lightweight convolutions, and we consider multiple new ways of integrating convolutions into Transformer self-attention. Specifically, we propose composite attention, which unites previous relative position embedding methods under a convolutional framework. We conduct experiments by training BERT with composite attention, finding that convolutions consistently improve performance on multiple downstream tasks, replacing absolute position embeddings. To inform future work, we present results comparing lightweight convolutions, dynamic convolutions, and depthwise-separable convolutions in language model pre-training, considering multiple injection points for convolutions in self-attention layers.

arXiv:2106.05482

Deep Position-wise Interaction Network for CTR Prediction

Click-through rate (CTR) prediction plays an important role in online advertising and recommender systems. In practice, the training of CTR models depends on click data which is intrinsically biased towards higher positions since higher position has higher CTR by nature. Existing methods such as actual position training with fixed position inference and inverse propensity weighted training with no position inference alleviate the bias problem to some extend. However, the different treatment of position information between training and inference will inevitably lead to inconsistency and sub-optimal online performance. Meanwhile, the basic assumption of these methods, i.e., the click probability is the product of examination probability and relevance probability, is oversimplified and insufficient to model the rich interaction between position and other information. In this paper, we propose a Deep Position-wise Interaction Network (DPIN) to efficiently combine all candidate items and positions for estimating CTR at each position, achieving consistency between offline and online as well as modeling the deep non-linear interaction among position, user, context and item under the limit of serving performance. Following our new treatment to the position bias in CTR prediction, we propose a new evaluation metrics named PAUC (position-wise AUC) that is suitable for measuring the ranking quality at a given position. Through extensive experiments on a real world dataset, we show empirically that our method is both effective and efficient in solving position bias problem. We have also deployed our method in production and observed statistically significant improvement over a highly optimized baseline in a rigorous A/B test.

arXiv:2106.05047공식 코드 · Papers with Code

Salient Object Ranking with Position-Preserved Attention

Instance segmentation can detect where the objects are in an image, but hard to understand the relationship between them. We pay attention to a typical relationship, relative saliency. A closely related task, salient object detection, predicts a binary map highlighting a visually salient region while hard to distinguish multiple objects. Directly combining two tasks by post-processing also leads to poor performance. There is a lack of research on relative saliency at present, limiting the practical applications such as content-aware image cropping, video summary, and image labeling. In this paper, we study the Salient Object Ranking (SOR) task, which manages to assign a ranking order of each detected object according to its visual saliency. We propose the first end-to-end framework of the SOR task and solve it in a multi-task learning fashion. The framework handles instance segmentation and salient object ranking simultaneously. In this framework, the SOR branch is independent and flexible to cooperate with different detection methods, so that easy to use as a plugin. We also introduce a Position-Preserved Attention (PPA) module tailored for the SOR branch. It consists of the position embedding stage and feature interaction stage. Considering the importance of position in saliency comparison, we preserve absolute coordinates of objects in ROI pooling operation and then fuse positional information with semantic features in the first stage. In the feature interaction stage, we apply the attention mechanism to obtain proposals' contextualized representations to predict their relative ranking orders. Extensive experiments have been conducted on the ASR dataset. Without bells and whistles, our proposed method outperforms the former state-of-the-art method significantly. The code will be released publicly available.

arXiv:2106.05022

Adaptive Inference through Early-Exit Networks: Design, Challenges and Directions

DNNs are becoming less and less over-parametrised due to recent advances in efficient model design, through careful hand-crafted or NAS-based methods. Relying on the fact that not all inputs require the same amount of computation to yield a confident prediction, adaptive inference is gaining attention as a prominent approach for pushing the limits of efficient deployment. Particularly, early-exit networks comprise an emerging direction for tailoring the computation depth of each input sample at runtime, offering complementary performance gains to other efficiency optimisations. In this paper, we decompose the design methodology of early-exit networks to its key components and survey the recent advances in each one of them. We also position early-exiting against other efficient inference solutions and provide our insights on the current challenges and most promising future directions for research in the field.

arXiv:2106.04263공식 코드 · Papers with Code

On the Connection between Local Attention and Dynamic Depth-wise Convolution

Vision Transformer (ViT) attains state-of-the-art performance in visual recognition, and the variant, Local Vision Transformer, makes further improvements. The major component in Local Vision Transformer, local attention, performs the attention separately over small local windows. We rephrase local attention as a channel-wise locally-connected layer and analyze it from two network regularization manners, sparse connectivity and weight sharing, as well as weight computation. Sparse connectivity: there is no connection across channels, and each position is connected to the positions within a small local window. Weight sharing: the connection weights for one position are shared across channels or within each group of channels. Dynamic weight: the connection weights are dynamically predicted according to each image instance. We point out that local attention resembles depth-wise convolution and its dynamic version in sparse connectivity. The main difference lies in weight sharing - depth-wise convolution shares connection weights (kernel weights) across spatial positions. We empirically observe that the models based on depth-wise convolution and the dynamic variant with lower computation complexity perform on-par with or sometimes slightly better than Swin Transformer, an instance of Local Vision Transformer, for ImageNet classification, COCO object detection and ADE semantic segmentation. These observations suggest that Local Vision Transformer takes advantage of two regularization forms and dynamic weight to increase the network capacity. Code is available at https://github.com/Atten4Vis/DemystifyLocalViT.

arXiv:2106.04306

Residual Feedback Learning for Contact-Rich Manipulation Tasks with Uncertainty

While classic control theory offers state of the art solutions in many problem scenarios, it is often desired to improve beyond the structure of such solutions and surpass their limitations. To this end, residual policy learning (RPL) offers a formulation to improve existing controllers with reinforcement learning (RL) by learning an additive "residual" to the output of a given controller. However, the applicability of such an approach highly depends on the structure of the controller. Often, internal feedback signals of the controller limit an RL algorithm to adequately change the policy and, hence, learn the task. We propose a new formulation that addresses these limitations by also modifying the feedback signals to the controller with an RL policy and show superior performance of our approach on a contact-rich peg-insertion task under position and orientation uncertainty. In addition, we use a recent Cartesian impedance control architecture as the control framework which can be available to us as a black-box while assuming no knowledge about its input/output structure, and show the difficulties of standard RPL. Furthermore, we introduce an adaptive curriculum for the given task to gradually increase the task difficulty in terms of position and orientation uncertainty. A video showing the results can be found at https://youtu.be/SAZm_Krze7U .

arXiv:2106.02795

Learnable Fourier Features for Multi-Dimensional Spatial Positional Encoding

Attentional mechanisms are order-invariant. Positional encoding is a crucial component to allow attention-based deep model architectures such as Transformer to address sequences or images where the position of information matters. In this paper, we propose a novel positional encoding method based on learnable Fourier features. Instead of hard-coding each position as a token or a vector, we represent each position, which can be multi-dimensional, as a trainable encoding based on learnable Fourier feature mapping, modulated with a multi-layer perceptron. The representation is particularly advantageous for a spatial multi-dimensional position, e.g., pixel positions on an image, where $L_2$ distances or more complex positional relationships need to be captured. Our experiments based on several public benchmark tasks show that our learnable Fourier feature representation for multi-dimensional positional encoding outperforms existing methods by both improving the accuracy and allowing faster convergence.

arXiv:2106.03518공식 코드 · Papers with Code

Position Bias Mitigation: A Knowledge-Aware Graph Model for Emotion Cause Extraction

The Emotion Cause Extraction (ECE)} task aims to identify clauses which contain emotion-evoking information for a particular emotion expressed in text. We observe that a widely-used ECE dataset exhibits a bias that the majority of annotated cause clauses are either directly before their associated emotion clauses or are the emotion clauses themselves. Existing models for ECE tend to explore such relative position information and suffer from the dataset bias. To investigate the degree of reliance of existing ECE models on clause relative positions, we propose a novel strategy to generate adversarial examples in which the relative position information is no longer the indicative feature of cause clauses. We test the performance of existing models on such adversarial examples and observe a significant performance drop. To address the dataset bias, we propose a novel graph-based method to explicitly model the emotion triggering paths by leveraging the commonsense knowledge to enhance the semantic dependencies between a candidate clause and an emotion clause. Experimental results show that our proposed approach performs on par with the existing state-of-the-art methods on the original ECE dataset, and is more robust against adversarial attacks compared to existing models.