arXiv subjects

Yilun Chen

Publications and source records attributed to Yilun Chen.

80 records · Page 5Linked to original sources

arXiv2019

Fast Point R-CNN

We present a unified, efficient and effective framework for point-cloud based 3D object detection. Our two-stage approach utilizes both voxel representation and raw point cloud data to exploit respective advantages. The first stage network, with voxel representation as input, only consists of light convolutional operations, producing a small number of high-quality initial predictions. Coordinate and indexed convolutional feature of each point in initial prediction are effectively fused with the attention mechanism, preserving both accurate localization and context information. The second stage works on interior points with their fused feature for further refining the prediction. Our method is evaluated on KITTI dataset, in terms of both 3D and Bird's Eye View (BEV) detection, and achieves state-of-the-arts with a 15FPS detection rate.

cs.CV↗

arXiv2018

Learning On-Road Visual Control for Self-Driving Vehicles with Auxiliary Tasks

A safe and robust on-road navigation system is a crucial component of achieving fully automated vehicles. NVIDIA recently proposed an End-to-End algorithm that can directly learn steering commands from raw pixels of a front camera by using one convolutional neural network. In this paper, we leverage auxiliary information aside from raw images and design a novel network structure, called Auxiliary Task Network (ATN), to help boost the driving performance while maintaining the advantage of minimal training data and an End-to-End training method. In this network, we introduce human prior knowledge into vehicle navigation by transferring features from image recognition tasks. Image semantic segmentation is applied as an auxiliary task for navigation. We consider temporal information by introducing an LSTM module and optical flow to the network. Finally, we combine vehicle kinematics with a sensor fusion step. We discuss the benefits of our method over state-of-the-art visual navigation methods both in the Udacity simulation environment and on the real-world Comma.ai dataset.

cs.CV↗

arXiv2018

Cascaded Pyramid Network for Multi-Person Pose Estimation

The topic of multi-person pose estimation has been largely improved recently, especially with the development of convolutional neural network. However, there still exist a lot of challenging cases, such as occluded keypoints, invisible keypoints and complex background, which cannot be well addressed. In this paper, we present a novel network structure called Cascaded Pyramid Network (CPN) which targets to relieve the problem from these "hard" keypoints. More specifically, our algorithm includes two stages: GlobalNet and RefineNet. GlobalNet is a feature pyramid network which can successfully localize the "simple" keypoints like eyes and hands but may fail to precisely recognize the occluded or invisible keypoints. Our RefineNet tries explicitly handling the "hard" keypoints by integrating all levels of feature representations from the GlobalNet together with an online hard keypoint mining loss. In general, to address the multi-person pose estimation problem, a top-down pipeline is adopted to first generate a set of human bounding boxes based on a detector, followed by our CPN for keypoint localization in each human bounding box. Based on the proposed algorithm, we achieve state-of-art results on the COCO keypoint benchmark, with average precision at 73.0 on the COCO test-dev dataset and 72.1 on the COCO test-challenge dataset, which is a 19% relative improvement compared with 60.5 from the COCO 2016 keypoint challenge.Code (https://github.com/chenyilun95/tf-cpn.git) and the detection results are publicly available for further research.

cs.CV↗

arXiv2013

Revealing social networks of spammers through spectral clustering

To date, most studies on spam have focused only on the spamming phase of the spam cycle and have ignored the harvesting phase, which consists of the mass acquisition of email addresses. It has been observed that spammers conceal their identity to a lesser degree in the harvesting phase, so it may be possible to gain new insights into spammers' behavior by studying the behavior of harvesters, which are individuals or bots that collect email addresses. In this paper, we reveal social networks of spammers by identifying communities of harvesters with high behavioral similarity using spectral clustering. The data analyzed was collected through Project Honey Pot, a distributed system for monitoring harvesting and spamming. Our main findings are (1) that most spammers either send only phishing emails or no phishing emails at all, (2) that most communities of spammers also send only phishing emails or no phishing emails at all, and (3) that several groups of spammers within communities exhibit coherent temporal behavior and have similar IP addresses. Our findings reveal some previously unknown behavior of spammers and suggest that there is indeed social structure between spammers to be discovered.

cs.SI↗

arXiv2011

Recursive $\ell_{1,\infty}$ Group lasso

We introduce a recursive adaptive group lasso algorithm for real-time penalized least squares prediction that produces a time sequence of optimal sparse predictor coefficient vectors. At each time index the proposed algorithm computes an exact update of the optimal $\ell_{1,\infty}$-penalized recursive least squares (RLS) predictor. Each update minimizes a convex but nondifferentiable function optimization problem. We develop an online homotopy method to reduce the computational complexity. Numerical simulations demonstrate that the proposed algorithm outperforms the $\ell_1$ regularized RLS algorithm for a group sparse system identification problem and has lower implementation complexity than direct group lasso solvers.

stat.ME↗

arXiv2010

Regularized Least-Mean-Square Algorithms

We consider adaptive system identification problems with convex constraints and propose a family of regularized Least-Mean-Square (LMS) algorithms. We show that with a properly selected regularization parameter the regularized LMS provably dominates its conventional counterpart in terms of mean square deviations. We establish simple and closed-form expressions for choosing this regularization parameter. For identifying an unknown sparse system we propose sparse and group-sparse LMS algorithms, which are special examples of the regularized LMS family. Simulation results demonstrate the advantages of the proposed filters in both convergence rate and steady-state error under sparsity assumptions on the true coefficient vector.

stat.ME↗

arXiv2010

Robust Shrinkage Estimation of High-dimensional Covariance Matrices

We address high dimensional covariance estimation for elliptical distributed samples, which are also known as spherically invariant random vectors (SIRV) or compound-Gaussian processes. Specifically we consider shrinkage methods that are suitable for high dimensional problems with a small number of samples (large $p$ small $n$). We start from a classical robust covariance estimator [Tyler(1987)], which is distribution-free within the family of elliptical distribution but inapplicable when $n<p$. Using a shrinkage coefficient, we regularize Tyler's fixed point iterations. We prove that, for all $n$ and $p$, the proposed fixed point iterations converge to a unique limit regardless of the initial condition. Next, we propose a simple, closed-form and data dependent choice for the shrinkage coefficient, which is based on a minimum mean squared error framework. Simulations demonstrate that the proposed method achieves low estimation error and is robust to heavy-tailed samples. Finally, as a real world application we demonstrate the performance of the proposed technique in the context of activity/intrusion detection using a wireless sensor network.

stat.ME↗

arXiv2009

Shrinkage Algorithms for MMSE Covariance Estimation

We address covariance estimation in the sense of minimum mean-squared error (MMSE) for Gaussian samples. Specifically, we consider shrinkage methods which are suitable for high dimensional problems with a small number of samples (large p small n). First, we improve on the Ledoit-Wolf (LW) method by conditioning on a sufficient statistic. By the Rao-Blackwell theorem, this yields a new estimator called RBLW, whose mean-squared error dominates that of LW for Gaussian variables. Second, to further reduce the estimation error, we propose an iterative approach which approximates the clairvoyant shrinkage estimator. Convergence of this iterative method is established and a closed form expression for the limit is determined, which is referred to as the oracle approximating shrinkage (OAS) estimator. Both RBLW and OAS estimators have simple expressions and are easily implemented. Although the two methods are developed from different persepctives, their structure is identical up to specified constants. The RBLW estimator provably dominates the LW method. Numerical simulations demonstrate that the OAS approach can perform even better than RBLW, especially when n is much less than p. We also demonstrate the performance of these techniques in the context of adaptive beamforming.

stat.ME↗