Searcharxiv⌕ Search

arXiv subjects

Dong Liu

Publications and source records attributed to Dong Liu.

At least 289 records · Page 16Linked to original sources
arXiv2021

Robust Classification using Hidden Markov Models and Mixtures of Normalizing Flows

We test the robustness of a maximum-likelihood (ML) based classifier where sequential data as observation is corrupted by noise. The hypothesis is that a generative model, that combines the state transitions of a hidden Markov model (HMM) and the neural network based probability distributions for the hidden states of the HMM, can provide a robust classification performance. The combined model is called normalizing-flow mixture model based HMM (NMM-HMM). It can be trained using a combination of expectation-maximization (EM) and backpropagation. We verify the improved robustness of NMM-HMM classifiers in an application to speech recognition.

cs.LG↗
arXiv2021

Learning and correcting non-Gaussian model errors

All discretized numerical models contain modelling errors - this reality is amplified when reduced-order models are used. The ability to accurately approximate modelling errors informs statistics on model confidence and improves quantitative results from frameworks using numerical models in prediction, tomography, and signal processing. Further to this, the compensation of highly nonlinear and non-Gaussian modelling errors, arising in many ill-conditioned systems aiming to capture complex physics, is a historically difficult task. In this work, we address this challenge by proposing a neural network approach capable of accurately approximating and compensating for such modelling errors in augmented direct and inverse problems. The viability of the approach is demonstrated using simulated and experimental data arising from differing physical direct and inverse problems.

physics.comp-ph↗
arXiv2020

Propagation of a plane-strain hydraulic fracture accounting for a rough cohesive zone

The quasi-brittle nature of rocks challenges the basic assumptions of linear hydraulic fracture mechanics (LHFM): linear elastic fracture mechanics and smooth parallel plates lubrication fluid flow. We relax these hypotheses and investigate the growth of a plane-strain hydraulic fracture in an impermeable medium accounting for a rough cohesive zone and a fluid lag. In addition to a dimensionless toughness and the time-scale of coalescence of the fluid and fracture fronts as in the LHFM case, the solution now also depends on the in-situ-to-cohesive stress ratio and the intensity of the flow deviation induced by aperture roughness. The solution is appropriately described by a nucleation time-scale, which delineates the fracture growth into a nucleation phase, an intermediate stage and a late time stage where convergence toward LHFM predictions finally occurs. A highly non-linear hydro-mechanical coupling takes place as the fluid front enters the rough cohesive zone which itself evolves during the nucleation and intermediate stages. This coupling leads to significant additional viscous flow dissipation. As a result, the fracture evolution deviates from LHFM solutions with shorter fracture lengths, larger widths and net pressures. These deviations ultimately decrease at late times as the lag and cohesive zone fractions both become smaller. The deviations increase with larger dimensionless toughness and in-situ-to-cohesive stress ratio, as both further localize viscous dissipation near the fluid front located in the rough cohesive zone. The convergence toward LHFM can occur at very late time for realistic values of in-situ-to-cohesive stress ratio encountered at depth. The impact of a rough cohesive zone appears to be prominent for laboratory experiments and short in-situ injections in quasi-brittle rocks with ultimately a larger energy demand compared to LHFM predictions.

physics.flu-dyn↗
arXiv2020

Accelerating Deep Reinforcement Learning With the Aid of Partial Model: Energy-Efficient Predictive Video Streaming

Predictive power allocation is conceived for energy-efficient video streaming over mobile networks using deep reinforcement learning. The goal is to minimize the accumulated energy consumption of each base station over a complete video streaming session under the constraint that avoids video playback interruptions. To handle the continuous state and action spaces, we resort to deep deterministic policy gradient (DDPG) algorithm for solving the formulated problem. In contrast to previous predictive power allocation policies that first predict future information with historical data and then optimize the power allocation based on the predicted information, the proposed policy operates in an on-line and end-to-end manner. By judiciously designing the action and state that only depend on slowly-varying average channel gains, we reduce the signaling overhead between the edge server and the base stations, and make it easier to learn a good policy. To further avoid playback interruption throughout the learning process and improve the convergence speed, we exploit the partially known model of the system dynamics by integrating the concepts of safety layer, post-decision state, and virtual experiences into the basic DDPG algorithm. Our simulation results show that the proposed policies converge to the optimal policy that is derived based on perfect large-scale channel prediction and outperform the first-predict-then-optimize policy in the presence of prediction errors. By harnessing the partially known model, the convergence speed can be dramatically improved.

cs.LG↗
arXiv2020

Vibro-acoustic baseline-free nondestructive testing

Vibro-Acoustic Modulation method for detection and characterization of various structural and material flaws has been actively researched for the last two decades. Most of the studies focused on detection and monitoring of macro-cracks requiring well established baseline (no-damage) value of the modulation index. The baseline value is specific for a particular structure, measuring setup, and other factors and can't be established in many practical situations without a long term monitoring looking for a relative change in the Modulation Index. In this work, we propose and investigate a baseline-free Vibro-Acoustic Modulation method, which does not require monitoring of relative Modulation Index change, unlike conventional approach. It was hypothesized that the nonlinear mechanisms (and respective nonlinear response) of a structure are different for undamaged and damaged material. For example: material without damage or at early stages of fatigue have classic elastic or hysteretic/dissipative nonlinearity while damaged (cracked) material may exhibit contact bi-linear or Hertzian nonlinear mechanisms. These mechanisms yield different power law dependencies of Modulation Index (MI) as function of applied vibration frequency input amplitude, B: MI ~ B^b Thus, quadratic nonlinearity yields linear dependence, b =1, and Hertzian nonlinearity results in b<1. Other nonlinear mechanisms yield different power laws. Therefore, measuring power damage coefficient b instead of MI may offer testing without established reference value. It is also offer some insights into the nonlinear mechanisms transformation during damage evolution. This approach was experimentally investigated and validated.

physics.app-ph↗
arXiv2020

Minimum-Delay Routing for Integrated Aeronautical Ad Hoc Networks Relying on Passenger-Planes in the North-Atlantic Region

Relying on multi-hop communication techniques, aeronautical ad hoc networks (AANETs) seamlessly integrate ground base stations (BSs) and satellites into aircraft communications for enhancing the on-demand connectivity of planes in the air. In this integrated AANET context we investigate the shortest-path routing problem with the objective of minimizing the total delay of the in-flight connection from the ground BS subject to certain minimum-rate constraints for all selected links in support of low-latency and high-speed services. Inspired by the best-first search and priority queue concepts, we model the problem formulated by a weighted digraph and find the optimal route based on the shortest-path algorithm. Our simulation results demonstrate that aircraft-aided multi-hop communications are capable of reducing the total delay of satellite communications, when relying on real historical flight data.

cs.IT↗
arXiv2020

Learned Variable-Rate Multi-Frequency Image Compression using Modulated Generalized Octave Convolution

In this proposal, we design a learned multi-frequency image compression approach that uses generalized octave convolutions to factorize the latent representations into high-frequency (HF) and low-frequency (LF) components, and the LF components have lower resolution than HF components, which can improve the rate-distortion performance, similar to wavelet transform. Moreover, compared to the original octave convolution, the proposed generalized octave convolution (GoConv) and octave transposed-convolution (GoTConv) with internal activation layers preserve more spatial structure of the information, and enable more effective filtering between the HF and LF components, which further improve the performance. In addition, we develop a variable-rate scheme using the Lagrangian parameter to modulate all the internal feature maps in the auto-encoder, which allows the scheme to achieve the large bitrate range of the JPEG AI with only three models. Experiments show that the proposed scheme achieves much better Y MS-SSIM than VVC. In terms of YUV PSNR, our scheme is very similar to HEVC.

eess.IV↗
arXiv2020

A Game Theoretic Analysis of LQG Control under Adversarial Attack

Motivated by recent works addressing adversarial attacks on deep reinforcement learning, a deception attack on linear quadratic Gaussian control is studied in this paper. In the considered attack model, the adversary can manipulate the observation of the agent subject to a mutual information constraint. The adversarial problem is formulated as a novel dynamic cheap talk game to capture the strategic interaction between the adversary and the agent, the asymmetry of information availability, and the system dynamics. Necessary and sufficient conditions are provided for subgame perfect equilibria to exist in pure strategies and in behavioral strategies; and characteristics of the equilibria and the resulting control rewards are given. The results show that pure strategy equilibria are informative, while only babbling equilibria exist in behavioral strategies. Numerical results are shown to illustrate the impact of strategic adversarial interaction.

eess.SY↗
arXiv2020

Learning Trailer Moments in Full-Length Movies

A movie's key moments stand out of the screenplay to grab an audience's attention and make movie browsing efficient. But a lack of annotations makes the existing approaches not applicable to movie key moment detection. To get rid of human annotations, we leverage the officially-released trailers as the weak supervision to learn a model that can detect the key moments from full-length movies. We introduce a novel ranking network that utilizes the Co-Attention between movies and trailers as guidance to generate the training pairs, where the moments highly corrected with trailers are expected to be scored higher than the uncorrelated moments. Additionally, we propose a Contrastive Attention module to enhance the feature representations such that the comparative contrast between features of the key and non-key moments are maximized. We construct the first movie-trailer dataset, and the proposed Co-Attention assisted ranking network shows superior performance even over the supervised approach. The effectiveness of our Contrastive Attention module is also demonstrated by the performance improvement over the state-of-the-art on the public benchmarks.

cs.CV↗
arXiv2020

Dual-Path Transformer Network: Direct Context-Aware Modeling for End-to-End Monaural Speech Separation

The dominant speech separation models are based on complex recurrent or convolution neural network that model speech sequences indirectly conditioning on context, such as passing information through many intermediate states in recurrent neural network, leading to suboptimal separation performance. In this paper, we propose a dual-path transformer network (DPTNet) for end-to-end speech separation, which introduces direct context-awareness in the modeling for speech sequences. By introduces a improved transformer, elements in speech sequences can interact directly, which enables DPTNet can model for the speech sequences with direct context-awareness. The improved transformer in our approach learns the order information of the speech sequences without positional encodings by incorporating a recurrent neural network into the original transformer. In addition, the structure of dual paths makes our model efficient for extremely long speech sequence modeling. Extensive experiments on benchmark datasets show that our approach outperforms the current state-of-the-arts (20.6 dB SDR on the public WSj0-2mix data corpus).

eess.AS↗
arXiv2020

Graph Neural Networks for Massive MIMO Detection

In this paper, we innovately use graph neural networks (GNNs) to learn a message-passing solution for the inference task of massive multiple multiple-input multiple-output (MIMO) detection in wireless communication. We adopt a graphical model based on the Markov random field (MRF) where belief propagation (BP) yields poor results when it assumes a uniform prior over the transmitted symbols. Numerical simulations show that, under the uniform prior assumption, our GNN-based MIMO detection solution outperforms the minimum mean-squared error (MMSE) baseline detector, in contrast to BP. Furthermore, experiments demonstrate that the performance of the algorithm slightly improves by incorporating MMSE information into the prior.

eess.SP↗
arXiv2020

Bottom-Up Human Pose Estimation by Ranking Heatmap-Guided Adaptive Keypoint Estimates

The typical bottom-up human pose estimation framework includes two stages, keypoint detection and grouping. Most existing works focus on developing grouping algorithms, e.g., associative embedding, and pixel-wise keypoint regression that we adopt in our approach. We present several schemes that are rarely or unthoroughly studied before for improving keypoint detection and grouping (keypoint regression) performance. First, we exploit the keypoint heatmaps for pixel-wise keypoint regression instead of separating them for improving keypoint regression. Second, we adopt a pixel-wise spatial transformer network to learn adaptive representations for handling the scale and orientation variance to further improve keypoint regression quality. Last, we present a joint shape and heatvalue scoring scheme to promote the estimated poses that are more likely to be true poses. Together with the tradeoff heatmap estimation loss for balancing the background and keypoint pixels and thus improving heatmap estimation quality, we get the state-of-the-art bottom-up human pose estimation result. Code is available at https://github.com/HRNet/HRNet-Bottom-up-Pose-Estimation.

cs.CV↗
arXiv2020

$α$ Belief Propagation for Approximate Inference

Belief propagation (BP) algorithm is a widely used message-passing method for inference in graphical models. BP on loop-free graphs converges in linear time. But for graphs with loops, BP's performance is uncertain, and the understanding of its solution is limited. To gain a better understanding of BP in general graphs, we derive an interpretable belief propagation algorithm that is motivated by minimization of a localized $α$-divergence. We term this algorithm as $α$ belief propagation ($α$-BP). It turns out that $α$-BP generalizes standard BP. In addition, this work studies the convergence properties of $α$-BP. We prove and offer the convergence conditions for $α$-BP. Experimental simulations on random graphs validate our theoretical results. The application of $α$-BP to practical problems is also demonstrated.

stat.ML↗
arXiv2020

Efficient Integer-Arithmetic-Only Convolutional Neural Networks

Integer-arithmetic-only networks have been demonstrated effective to reduce computational cost and to ensure cross-platform consistency. However, previous works usually report a decline in the inference accuracy when converting well-trained floating-point-number (FPN) networks into integer networks. We analyze this phonomenon and find that the decline is due to activation quantization. Specifically, when we replace conventional ReLU with Bounded ReLU, how to set the bound for each neuron is a key problem. Considering the tradeoff between activation quantization error and network learning ability, we set an empirical rule to tune the bound of each Bounded ReLU. We also design a mechanism to handle the cases of feature map addition and feature map concatenation. Based on the proposed method, our trained 8-bit integer ResNet outperforms the 8-bit networks of Google's TensorFlow and NVIDIA's TensorRT for image recognition. We also experiment on VDSR for image super-resolution and on VRCNN for compression artifact reduction, both of which serve for regression tasks that natively require high inference accuracy. Our integer networks achieve equivalent performance as the corresponding FPN networks, but have only 1/4 memory cost and run 2x faster on modern GPUs. Our code and models can be found at github.com/HengRuiZ/brelu.

cs.CV↗
arXiv2020

Region-based Energy Neural Network for Approximate Inference

Region-based free energy was originally proposed for generalized belief propagation (GBP) to improve loopy belief propagation (loopy BP). In this paper, we propose a neural network based energy model for inference in general Markov random fields (MRFs), which directly minimizes the region-based free energy defined on region graphs. We term our model Region-based Energy Neural Network (RENN). Unlike message-passing algorithms, RENN avoids iterative message propagation and is faster. Also different from recent deep neural network based models, inference by RENN does not require sampling, and RENN works on general MRFs. RENN can also be employed for MRF learning. Our experiments on marginal distribution estimation, partition function estimation, and learning of MRFs show that RENN outperforms the mean field method, loopy BP, GBP, and the state-of-the-art neural network based model.

cs.LG↗
arXiv2020

Foreground-Background Imbalance Problem in Deep Object Detectors: A Review

Recent years have witnessed the remarkable developments made by deep learning techniques for object detection, a fundamentally challenging problem of computer vision. Nevertheless, there are still difficulties in training accurate deep object detectors, one of which is owing to the foreground-background imbalance problem. In this paper, we survey the recent advances about the solutions to the imbalance problem. First, we analyze the characteristics of the imbalance problem in different kinds of deep detectors, including one-stage and two-stage ones. Second, we divide the existing solutions into two categories: sampling heuristics and non-sampling schemes, and review them in detail. Third, we experimentally compare the performance of some state-of-the-art solutions on the COCO benchmark. Promising directions for future work are also discussed.

cs.CV↗