arXiv subjects

Xiang Chen

Publications and source records attributed to Xiang Chen.

At least 505 records · Page 28Linked to original sources

arXiv2020

Generative Feature Replay with Orthogonal Weight Modification for Continual Learning

The ability of intelligent agents to learn and remember multiple tasks sequentially is crucial to achieving artificial general intelligence. Many continual learning (CL) methods have been proposed to overcome catastrophic forgetting which results from non i.i.d data in the sequential learning of neural networks. In this paper we focus on class incremental learning, a challenging CL scenario. For this scenario, generative replay is a promising strategy which generates and replays pseudo data for previous tasks to alleviate catastrophic forgetting. However, it is hard to train a generative model continually for relatively complex data. Based on recently proposed orthogonal weight modification (OWM) algorithm which can approximately keep previously learned feature invariant when learning new tasks, we propose to 1) replay penultimate layer feature with a generative model; 2) leverage a self-supervised auxiliary task to further enhance the stability of feature. Empirical results on several datasets show our method always achieves substantial improvement over powerful OWM while conventional generative replay always results in a negative effect. Meanwhile our method beats several strong baselines including one based on real data storage. In addition, we conduct experiments to study why our method is effective.

cs.LG↗

arXiv2020

Unconventional Hund Metal in a Weak Itinerant Ferromagnet

The physics of weak itinerant ferromagnets is challenging due to their small magnetic moments and the ambiguous role of local interactions governing their electronic properties, many of which violate Fermi liquid theory. While magnetic fluctuations play an important role in the materials' unusual electronic states, the nature of these fluctuations and the paradigms through which they arise remain debated. Here we use inelastic neutron scattering to study magnetic fluctuations in the canonical weak itinerant ferromagnet MnSi. Data reveal that short-wavelength magnons continue to propagate until a mode crossing predicted for strongly interacting quasiparticles is reached, and the local susceptibility peaks at a coherence energy predicted for a correlated Hund metal by first-principles many-body theory. Scattering between electrons and orbital and spin fluctuations in MnSi can be understood at the local level to generate non-Fermi liquid character. These results provide crucial insight into the role of interorbital Hund's exchange within the broader class of enigmatic multiband itinerant, weak ferromagnets.

cond-mat.str-el↗

arXiv2020

MLBF-Net: A Multi-Lead-Branch Fusion Network for Multi-Class Arrhythmia Classification Using 12-Lead ECG

Automatic arrhythmia detection using 12-lead electrocardiogram (ECG) signal plays a critical role in early prevention and diagnosis of cardiovascular diseases. In the previous studies on automatic arrhythmia detection, most methods concatenated 12 leads of ECG into a matrix, and then input the matrix to a variety of feature extractors or deep neural networks for extracting useful information. Under such frameworks, these methods had the ability to extract comprehensive features (known as integrity) of 12-lead ECG since the information of each lead interacts with each other during training. However, the diverse lead-specific features (known as diversity) among 12 leads were neglected, causing inadequate information learning for 12-lead ECG. To maximize the information learning of multi-lead ECG, the information fusion of comprehensive features with integrity and lead-specific features with diversity should be taken into account. In this paper, we propose a novel Multi-Lead-Branch Fusion Network (MLBF-Net) architecture for arrhythmia classification by integrating multi-loss optimization to jointly learning diversity and integrity of multi-lead ECG. MLBF-Net is composed of three components: 1) multiple lead-specific branches for learning the diversity of multi-lead ECG; 2) cross-lead features fusion by concatenating the output feature maps of all branches for learning the integrity of multi-lead ECG; 3) multi-loss co-optimization for all the individual branches and the concatenated network. We demonstrate our MLBF-Net on China Physiological Signal Challenge 2018 which is an open 12-lead ECG dataset. The experimental results show that MLBF-Net obtains an average $F_1$ score of 0.855, reaching the highest arrhythmia classification performance. The proposed method provides a promising solution for multi-lead ECG analysis from an information fusion perspective.

eess.SP↗

arXiv2020

AntiDote: Attention-based Dynamic Optimization for Neural Network Runtime Efficiency

Convolutional Neural Networks (CNNs) achieved great cognitive performance at the expense of considerable computation load. To relieve the computation load, many optimization works are developed to reduce the model redundancy by identifying and removing insignificant model components, such as weight sparsity and filter pruning. However, these works only evaluate model components' static significance with internal parameter information, ignoring their dynamic interaction with external inputs. With per-input feature activation, the model component significance can dynamically change, and thus the static methods can only achieve sub-optimal results. Therefore, we propose a dynamic CNN optimization framework in this work. Based on the neural network attention mechanism, we propose a comprehensive dynamic optimization framework including (1) testing-phase channel and column feature map pruning, as well as (2) training-phase optimization by targeted dropout. Such a dynamic optimization framework has several benefits: (1) First, it can accurately identify and aggressively remove per-input feature redundancy with considering the model-input interaction; (2) Meanwhile, it can maximally remove the feature map redundancy in various dimensions thanks to the multi-dimension flexibility; (3) The training-testing co-optimization favors the dynamic pruning and helps maintain the model accuracy even with very high feature pruning ratio. Extensive experiments show that our method could bring 37.4% to 54.5% FLOPs reduction with negligible accuracy drop on various of test networks.

cs.CV↗

arXiv2020

RapidLayout: Fast Hard Block Placement of FPGA-optimized Systolic Arrays using Evolutionary Algorithms

Evolutionary algorithms can outperform conventional placement algorithms such as simulated annealing, analytical placement as well as manual placement on metrics such as runtime, wirelength, pipelining cost, and clock frequency when mapping FPGA hard block intensive designs such as systolic arrays on Xilinx UltraScale+ FPGAs. For certain hard-block intensive, systolic array accelerator designs, the commercial-grade Xilinx Vivado CAD tool is unable to provide a legal routing solution without tedious manual placement constraints. Instead, we formulate an automatic FPGA placement algorithm for these hard blocks as a multi-objective optimization problem that targets wirelength squared and maximum bounding box size metrics. We build an end-to-end placement and routing flow called RapidLayout using the Xilinx RapidWright framework. RapidLayout runs 5-6$\times$ faster than Vivado with manual constraints and eliminates the weeks-long effort to generate placement constraints manually for the hard blocks. We also perform automated post-placement pipelining of the long wires inside each convolution block to target 650MHz URAM-limited operation. RapidLayout outperforms (1) the simulated annealer in VPR by 33% in runtime, 1.9-2.4$\times$ in wirelength, and 3-4$\times$ in bounding box size, while also (2) beating the analytical placer UTPlaceF by 9.3$\times$ in runtime, 1.8-2.2$\times$ in wirelength, and 2-2.7$\times$ in bounding box size. We employ transfer learning from a base FPGA device to speed-up placement optimization for similar FPGA devices in the UltraScale+ family by 11-14$\times$ than learning the placements from scratch.

cs.AR↗

arXiv2020

An Image Enhancing Pattern-based Sparsity for Real-time Inference on Mobile Devices

Weight pruning has been widely acknowledged as a straightforward and effective method to eliminate redundancy in Deep Neural Networks (DNN), thereby achieving acceleration on various platforms. However, most of the pruning techniques are essentially trade-offs between model accuracy and regularity which lead to impaired inference accuracy and limited on-device acceleration performance. To solve the problem, we introduce a new sparsity dimension, namely pattern-based sparsity that comprises pattern and connectivity sparsity, and becoming both highly accurate and hardware friendly. With carefully designed patterns, the proposed pruning unprecedentedly and consistently achieves accuracy enhancement and better feature extraction ability on different DNN structures and datasets, and our pattern-aware pruning framework also achieves pattern library extraction, pattern selection, pattern and connectivity pruning and weight training simultaneously. Our approach on the new pattern-based sparsity naturally fits into compiler optimization for highly efficient DNN execution on mobile platforms. To the best of our knowledge, it is the first time that mobile devices achieve real-time inference for the large-scale DNN models thanks to the unique spatial property of pattern-based sparsity and the help of the code generation capability of compilers.

cs.CV↗

arXiv2020

LR-CNN: Local-aware Region CNN for Vehicle Detection in Aerial Imagery

State-of-the-art object detection approaches such as Fast/Faster R-CNN, SSD, or YOLO have difficulties detecting dense, small targets with arbitrary orientation in large aerial images. The main reason is that using interpolation to align RoI features can result in a lack of accuracy or even loss of location information. We present the Local-aware Region Convolutional Neural Network (LR-CNN), a novel two-stage approach for vehicle detection in aerial imagery. We enhance translation invariance to detect dense vehicles and address the boundary quantization issue amongst dense vehicles by aggregating the high-precision RoIs' features. Moreover, we resample high-level semantic pooled features, making them regain location information from the features of a shallower convolutional block. This strengthens the local feature invariance for the resampled features and enables detecting vehicles in an arbitrary orientation. The local feature invariance enhances the learning ability of the focal loss function, and the focal loss further helps to focus on the hard examples. Taken together, our method better addresses the challenges of aerial imagery. We evaluate our approach on several challenging datasets (VEDAI, DOTA), demonstrating a significant improvement over state-of-the-art methods. We demonstrate the good generalization ability of our approach on the DLR 3K dataset.

cs.CV↗

arXiv2020

Radial Coverage Strength for Optimization of Multi-Camera Deployment

In this paper, a new concept, radial coverage strength, is first proposed to characterize the visual sensing performance when the orientation of the target pose is considered. In particular, the elevation angle of the optical pose of the visual sensor is taken to decompose the visual coverage strength into effective and ineffective components, motivated by the imaging intuition. An optimization problem is then formulated for a multi-camera network to maximize the coverage of the object area based on the strength information fusion along the effective coverage strength direction through the deployment of the angle between radial coverage vector of the camera optical pose. Both simulation and experiments are conducted to validate the proposed approach and comparison with existing methods is also provided.

eess.IV↗

arXiv2020

Average Age of Changed Information in the Internet of Things

The freshness of status updates is imperative in mission-critical Internet of things (IoT) applications. Recently, Age of Information (AoI) has been proposed to measure the freshness of updates at the receiver. However, AoI only characterizes the freshness over time, but ignores the freshness in the content. In this paper, we introduce a new performance metric, Age of Changed Information (AoCI), which captures both the passage of time and the change of information content. Also, we examine the AoCI in a time-slotted status update system, where a sensor samples the physical process and transmits the update packets with a cost. We formulate a Markov Decision Process (MDP) to find the optimal updating policy that minimizes the weighted sum of the AoCI and the update cost. Particularly, in a special case that the physical process is modeled by a two-state discrete time Markov chain with equal transition probability, we show that the optimal policy is of threshold type with respect to the AoCI and derive the closed-form of the threshold. Finally, simulations are conducted to exhibit the performance of the threshold policy and its superiority over the zero-wait baseline policy.

cs.IT↗

arXiv2020

Joint Transmission and Computing Scheduling for Status Update with Mobile Edge Computing

Age of Information (AoI), defined as the time elapsed since the generation of the latest received update, is a promising performance metric to measure data freshness for real-time status monitoring. In many applications, status information needs to be extracted through computing, which can be processed at an edge server enabled by mobile edge computing (MEC). In this paper, we aim to minimize the average AoI within a given deadline by jointly scheduling the transmissions and computations of a series of update packets with deterministic transmission and computing times. The main analytical results are summarized as follows. Firstly, the minimum deadline to guarantee the successful transmission and computing of all packets is given. Secondly, a \emph{no-wait computing} policy which intuitively attains the minimum AoI is introduced, and the feasibility condition of the policy is derived. Finally, a closed-form optimal scheduling policy is obtained on the condition that the deadline exceeds a certain threshold. The behavior of the optimal transmission and computing policy is illustrated by numerical results with different values of the deadline, which validates the analytical results.

cs.IT↗

arXiv2020

Analysis on Computation-Intensive Status Update in Mobile Edge Computing

In status update scenarios, the freshness of information is measured in terms of age-of-information (AoI), which essentially reflects the timeliness for real-time applications to transmit status update messages to a remote controller. For some applications, computational expensive and time consuming data processing is inevitable for status information of messages to be displayed. Mobile edge servers are equipped with adequate computation resources and they are placed close to users. Thus, mobile edge computing (MEC) can be a promising technology to reduce AoI for computation-intensive messages. In this paper, we study the AoI for computation-intensive messages with MEC, and consider three computing schemes: local computing, remote computing at the MEC server, and partial computing, i.e., some part of computing tasks are performed locally, and the rest is executed at the MEC server. Zero-wait policy is adopted in all three schemes. Specifically, in local computing, a new message is generated immediately after the previous one is revealed by computing. While in remote computing and partial computing, a new message is generated once the previous one is received by the remote MEC server. With infinite queue size and exponentially distributed transmission time, closed-form average AoI for exponentially distributed computing time is derived for the three computing schemes. For deterministic computing time, the average AoI is analyzed numerically. Simulation results show that by carefully partitioning the computing tasks, the average AoI in partial computing is the smallest compared to local computing and remote computing. The results also indicate numerically the conditions on which remote computing attains smaller average AoI compared with local computing.

cs.IT↗

arXiv2020

SQLFlow: A Bridge between SQL and Machine Learning

Industrial AI systems are mostly end-to-end machine learning (ML) workflows. A typical recommendation or business intelligence system includes many online micro-services and offline jobs. We describe SQLFlow for developing such workflows efficiently in SQL. SQL enables developers to write short programs focusing on the purpose (what) and ignoring the procedure (how). Previous database systems extended their SQL dialect to support ML. SQLFlow (https://sqlflow.org/sqlflow ) takes another strategy to work as a bridge over various database systems, including MySQL, Apache Hive, and Alibaba MaxCompute, and ML engines like TensorFlow, XGBoost, and scikit-learn. We extended SQL syntax carefully to make the extension working with various SQL dialects. We implement the extension by inventing a collaborative parsing algorithm. SQLFlow is efficient and expressive to a wide variety of ML techniques -- supervised and unsupervised learning; deep networks and tree models; visual model explanation in addition to training and prediction; data processing and feature extraction in addition to ML. SQLFlow compiles a SQL program into a Kubernetes-native workflow for fault-tolerable execution and on-cloud deployment. Current industrial users include Ant Financial, DiDi, and Alibaba Group.

cs.DB↗

arXiv2019

Spectroscopic evidence for electron-boson coupling in electron-doped Sr$_2$IrO$_4$

The pseudogap, d-wave superconductivity and electron-boson coupling are three intertwined key ingredients in the phase diagram of the cuprates. Sr$_2$IrO$_4$ is a 5d-electron counterpart of the cuprates in which both the pseudogap and a d-wave instability have been observed. Here, we report spectroscopic evidence for the presence of the third key player in electron-doped Sr$_2$IrO$_4$: electron-boson coupling. A kink in nodal dispersion is observed with an energy scale of ~50 meV. The strength of the kink changes with doping, but the energy scale remains the same. These results provide the first noncuprate platform for exploring the relationship between the pseudogap, d-wave instability and electron-boson coupling in doped Mott insulators.

cond-mat.supr-con↗

arXiv2019

Atomic-scale fragmentation and collapse of antiferromagnetic order in a doped Mott insulator

Disentangling the relationship between the insulating state with a charge gap and the magnetic order in an antiferromagnetic (AF) Mott insulator remains difficult due to inherent phase separation as the Mott state is perturbed. Measuring magnetic and electronic properties at the atomic length scales would provide crucial insight, but this is yet to be experimentally achieved. Here we use spectroscopic-imaging spin-polarized scanning tunneling microscopy (SP-STM) to visualize periodic spin-resolved modulations originating from the AF order in a relativistic Mott insulator Sr2IrO4, and study these as a function of doping. We find that near insulator-to-metal transition (IMT), the long-range AF order melts into a fragmented state with short-range AF correlations. Crucially, we discover that the short-range AF order is locally uncorrelated with the observed spectral gap magnitude. This strongly suggests that short range AF correlations are unlikely to be the culprit behind inhomogeneous gap closing and the emergence of pseudogap regions near IMT. Our work establishes SP-STM as a powerful tool for revealing atomic-scale magnetic information in complex oxides.

cond-mat.str-el↗

arXiv2019

LanCe: A Comprehensive and Lightweight CNN Defense Methodology against Physical Adversarial Attacks on Embedded Multimedia Applications

Recently, adversarial attacks can be applied to the physical world, causing practical issues to various Convolutional Neural Networks (CNNs) powered applications. Most existing physical adversarial attack defense works only focus on eliminating explicit perturbation patterns from inputs, ignoring interpretation to CNN's intrinsic vulnerability. Therefore, they lack the expected versatility to different attacks and thereby depend on considerable data processing costs. In this paper, we propose LanCe -- a comprehensive and lightweight CNN defense methodology against different physical adversarial attacks. By interpreting CNN's vulnerability, we find that non-semantic adversarial perturbations can activate CNN with significantly abnormal activations and even overwhelm other semantic input patterns' activations. We improve the CNN recognition process by adding a self-verification stage to detect the potential adversarial input with only one CNN inference cost. Based on the detection result, we further propose a data recovery methodology to defend the physical adversarial attacks. We apply such defense methodology into both image and audio CNN recognition scenarios and analyze the computational complexity for each scenario, respectively. Experiments show that our methodology can achieve an average 91% successful rate for attack detection and 89% accuracy recovery. Moreover, it is at most 3x faster compared with the state-of-the-art defense methods, making it feasible to resource-constrained embedded systems, such as mobile devices.

cs.CV↗

arXiv2019

Task-Adaptive Incremental Learning for Intelligent Edge Devices

Convolutional Neural Networks (CNNs) are used for a wide range of image-related tasks such as image classification and object detection. However, a large pre-trained CNN model contains a lot of redundancy considering the task-specific edge applications. Also, the statically pre-trained model could not efficiently handle the dynamic data in the real-world application. The CNN training data and their labels are collected in an incremental manner. To tackle the above two challenges, we proposed TeAM a task-adaptive incremental learning framework for CNNs in intelligent edge devices. Given a pre-trained large model, TeAM can configure it into any specialized model for dedicated edge applications. The specialized model can be quickly fine-tuned with local data to achieve very high accuracy. Also, with our global aggregation and incremental learning scheme, the specialized CNN models can be collaboratively aggregated to an enhanced global model with new training data.

cs.DC↗

arXiv2019

Functionality-Oriented Convolutional Filter Pruning

The sophisticated structure of Convolutional Neural Network (CNN) allows for outstanding performance, but at the cost of intensive computation. As significant redundancies inevitably present in such a structure, many works have been proposed to prune the convolutional filters for computation cost reduction. Although extremely effective, most works are based only on quantitative characteristics of the convolutional filters, and highly overlook the qualitative interpretation of individual filter's specific functionality. In this work, we interpreted the functionality and redundancy of the convolutional filters from different perspectives, and proposed a functionality-oriented filter pruning method. With extensive experiment results, we proved the convolutional filters' qualitative significance regardless of magnitude, demonstrated significant neural network redundancy due to repetitive filter functions, and analyzed the filter functionality defection under inappropriate retraining process. Such an interpretable pruning approach not only offers outstanding computation cost optimization over previous filter pruning methods, but also interprets filter pruning process.

cs.LG↗

arXiv2019

DoPa: A Comprehensive CNN Detection Methodology against Physical Adversarial Attacks

Recently, Convolutional Neural Networks (CNNs) demonstrate a considerable vulnerability to adversarial attacks, which can be easily misled by adversarial perturbations. With more aggressive methods proposed, adversarial attacks can be also applied to the physical world, causing practical issues to various CNN powered applications. To secure CNNs, adversarial attack detection is considered as the most critical approach. However, most existing works focus on superficial patterns and merely search a particular method to differentiate the adversarial inputs and natural inputs, ignoring the analysis of CNN inner vulnerability. Therefore, they can only target to specific physical adversarial attacks, lacking expected versatility to different attacks. To address this issue, we propose DoPa -- a comprehensive CNN detection methodology for various physical adversarial attacks. By interpreting the CNN's vulnerability, we find that non-semantic adversarial perturbations can activate CNN with significantly abnormal activations and even overwhelm other semantic input patterns' activations. Therefore, we add a self-verification stage to analyze the semantics of distinguished activation patterns, which improves the CNN recognition process. We apply such a detection methodology into both image and audio CNN recognition scenarios. Experiments show that DoPa can achieve an average rate of 90% success for image attack detection and 92% success for audio attack detection. Announcement:[The original DoPa draft on arXiv was modified and submitted to a conference already, while this short abstract was submitted only for a presentation at the KDD 2019 AIoT Workshop.]

cs.CR↗