arXiv subjects

Liu Liu

Publications and source records attributed to Liu Liu.

At least 271 records · Page 15Linked to original sources

arXiv2019

NODE: Extreme Low Light Raw Image Denoising using a Noise Decomposition Network

Denoising extreme low light images is a challenging task due to the high noise level. When the illumination is low, digital cameras increase the ISO (electronic gain) to amplify the brightness of captured data. However, this in turn amplifies the noise, arising from read, shot, and defective pixel sources. In the raw domain, read and shot noise are effectively modelled using Gaussian and Poisson distributions respectively, whereas defective pixels can be modeled with impulsive noise. In extreme low light imaging, noise removal becomes a critical challenge to produce a high quality, detailed image with low noise. In this paper, we propose a multi-task deep neural network called Noise Decomposition (NODE) that explicitly and separately estimates defective pixel noise, in conjunction with Gaussian and Poisson noise, to denoise an extreme low light image. Our network is purposely designed to work with raw data, for which the noise is more easily modeled before going through non-linear transformations in the image signal processing (ISP) pipeline. Quantitative and qualitative evaluation show the proposed method to be more effective at denoising real raw images than state-of-the-art techniques.

eess.IV↗

arXiv2019

A bi-fidelity method for the multiscale Boltzmann equation with random parameters

In this paper, we study the multiscale Boltzmann equation with multi-dimensional random parameters by a bi-fidelity stochastic collocation (SC) method developed in [A. Narayan, C. Gittelson and D. Xiu, SIAM J. Sci. Comput., 36 (2014); X. Zhu, A. Narayan and D. Xiu, SIAM J. Uncertain. Quantif., 2 (2014)]. By choosing the compressible Euler system as the low-fidelity model, we adapt the bi-fidelity SC method to combine computational efficiency of the low-fidelity model with high accuracy of the high-fidelity (Boltzmann) model. With only a small number of high-fidelity asymptotic-preserving solver runs for the Boltzmann equation, the bi-fidelity approximation can capture well the macroscopic quantities of the solution to the Boltzmann equation in the random space. A priori estimate on the accuracy between the high- and bi-fidelity solutions together with a convergence analysis is established. Finally, we present extensive numerical experiments to verify the efficiency and accuracy of our proposed method.

math.NA↗

arXiv2019

Asymptotic-preserving schemes for two-species binary collisional kinetic system with disparate masses I: time discretization and asymptotic analysis

We develop efficient asymptotic-preserving time discretization schemes to solve the disparate mass kinetic system of a binary gas or plasma in the "relaxation time scale" relevant to the epochal relaxation phenomenon. Since the resulting model is associated to a parameter given by the square of the mass ratio between the light and heavy particles, we develop an asymptotic-preserving scheme in a novel decomposition strategy using the penalization method for multiscale collisional kinetic equations. Both the Boltzmann and Fokker-Planck-Landau (FPL) binary collision operators will be considered. Other than utilizing several AP strategies for single-species binary kinetic equations, we also introduce a novel splitting and a carefully designed explicit-implicit approximation, which are guided by the asymptotic analysis of the system. We also conduct asymptotic-preserving analysis for the time discretization, for both space homogenous and inhomogeneous systems.

math.NA↗

arXiv2019

Stochastic Attraction-Repulsion Embedding for Large Scale Image Localization

This paper tackles the problem of large-scale image-based localization (IBL) where the spatial location of a query image is determined by finding out the most similar reference images in a large database. For solving this problem, a critical task is to learn discriminative image representation that captures informative information relevant for localization. We propose a novel representation learning method having higher location-discriminating power. It provides the following contributions: 1) we represent a place (location) as a set of exemplar images depicting the same landmarks and aim to maximize similarities among intra-place images while minimizing similarities among inter-place images; 2) we model a similarity measure as a probability distribution on L_2-metric distances between intra-place and inter-place image representations; 3) we propose a new Stochastic Attraction and Repulsion Embedding (SARE) loss function minimizing the KL divergence between the learned and the actual probability distributions; 4) we give theoretical comparisons between SARE, triplet ranking and contrastive losses. It provides insights into why SARE is better by analyzing gradients. Our SARE loss is easy to implement and pluggable to any CNN. Experiments show that our proposed method improves the localization performance on standard benchmarks by a large margin. Demonstrating the broad applicability of our method, we obtained the third place out of 209 teams in the 2018 Google Landmark Retrieval Challenge. Our code and model are available at https://github.com/Liumouliu/deepIBL.

cs.CV↗

arXiv2019

HODGEPODGE: Sound event detection based on ensemble of semi-supervised learning methods

In this paper, we present a method called HODGEPODGE\footnotemark[1] for large-scale detection of sound events using weakly labeled, synthetic, and unlabeled data proposed in the Detection and Classification of Acoustic Scenes and Events (DCASE) 2019 challenge Task 4: Sound event detection in domestic environments. To perform this task, we adopted the convolutional recurrent neural networks (CRNN) as our backbone network. In order to deal with a small amount of tagged data and a large amounts of unlabeled in-domain data, we aim to focus primarily on how to apply semi-supervise learning methods efficiently to make full use of limited data. Three semi-supervised learning principles have been used in our system, including: 1) Consistency regularization applies data augmentation; 2) MixUp regularizer requiring that the predictions for a interpolation of two inputs is close to the interpolation of the prediction for each individual input; 3) MixUp regularization applies to interpolation between data augmentations. We also tried an ensemble of various models, which are trained by using different semi-supervised learning principles. Our proposed approach significantly improved the performance of the baseline, achieving the event-based f-measure of 42.0\% compared to 25.8\% event-based f-measure of the baseline in the provided official evaluation dataset. Our submissions ranked third among 18 teams in the task 4.

cs.SD↗

arXiv2019

High Dimensional Robust Sparse Regression

We provide a novel -- and to the best of our knowledge, the first -- algorithm for high dimensional sparse regression with constant fraction of corruptions in explanatory and/or response variables. Our algorithm recovers the true sparse parameters with sub-linear sample complexity, in the presence of a constant fraction of arbitrary corruptions. Our main contribution is a robust variant of Iterative Hard Thresholding. Using this, we provide accurate estimators: when the covariance matrix in sparse regression is identity, our error guarantee is near information-theoretically optimal. We then deal with robust sparse regression with unknown structured covariance matrix. We propose a filtering algorithm which consists of a novel randomized outlier removal technique for robust sparse mean estimation that may be of interest in its own right: the filtering algorithm is flexible enough to deal with unknown covariance. Also, it is orderwise more efficient computationally than the ellipsoid algorithm. Using sub-linear sample complexity, our algorithm achieves the best known (and first) error guarantee. We demonstrate the effectiveness on large-scale sparse regression problems with arbitrary corruptions.

cs.LG↗

arXiv2019

High Dimensional Robust $M$-Estimation: Arbitrary Corruption and Heavy Tails

We consider the problem of sparsity-constrained $M$-estimation when both explanatory and response variables have heavy tails (bounded 4-th moments), or a fraction of arbitrary corruptions. We focus on the $k$-sparse, high-dimensional regime where the number of variables $d$ and the sample size $n$ are related through $n \sim k \log d$. We define a natural condition we call the Robust Descent Condition (RDC), and show that if a gradient estimator satisfies the RDC, then Robust Hard Thresholding (IHT using this gradient estimator), is guaranteed to obtain good statistical rates. The contribution of this paper is in showing that this RDC is a flexible enough concept to recover known results, and obtain new robustness results. Specifically, new results include: (a) For $k$-sparse high-dimensional linear- and logistic-regression with heavy tail (bounded 4-th moment) explanatory and response variables, a linear-time-computable median-of-means gradient estimator satisfies the RDC, and hence Robust Hard Thresholding is minimax optimal; (b) When instead of heavy tails we have $O(1/\sqrt{k}\log(nd))$-fraction of arbitrary corruptions in explanatory and response variables, a near linear-time computable trimmed gradient estimator satisfies the RDC, and hence Robust Hard Thresholding is minimax optimal. We demonstrate the effectiveness of our approach in sparse linear, logistic regression, and sparse precision matrix estimation on synthetic and real-world US equities data.

cs.LG↗

arXiv2019

Dynamic Sparse Graph for Efficient Deep Learning

We propose to execute deep neural networks (DNNs) with dynamic and sparse graph (DSG) structure for compressive memory and accelerative execution during both training and inference. The great success of DNNs motivates the pursuing of lightweight models for the deployment onto embedded devices. However, most of the previous studies optimize for inference while neglect training or even complicate it. Training is far more intractable, since (i) the neurons dominate the memory cost rather than the weights in inference; (ii) the dynamic activation makes previous sparse acceleration via one-off optimization on fixed weight invalid; (iii) batch normalization (BN) is critical for maintaining accuracy while its activation reorganization damages the sparsity. To address these issues, DSG activates only a small amount of neurons with high selectivity at each iteration via a dimension-reduction search (DRS) and obtains the BN compatibility via a double-mask selection (DMS). Experiments show significant memory saving (1.7-4.5x) and operation reduction (2.3-4.4x) with little accuracy loss on various benchmarks.

cs.LG↗

arXiv2019

A Polarization-insensitive and High-speed Electro-optic Switch Based on a Hybrid Silicon and Lithium Niobate Platform

We propose and demonstrate a polarization-insensitive and high speed optical switch unit based on a silicon and lithium niobate hybrid integration platform. The presented device exhibits a sub nano-second switching time, low drive voltages of 4.97 V, and low power dissipation due to electrostatic operation. The measured polarization dependence loss was lower than 0.8 dB. The demonstrated optical switch could provide as a building block for polarization-insensitive and high-speed optical matrix switches.

physics.app-ph↗

arXiv2019

Lithium Niobate Michelson Interferometer Modulator on Silicon-On-Insulator Platform

We propose and demonstrate a hybrid silicon and lithium niobate Michelson Interferometer Modulator (MIM) with enhanced modulation efficiency compared to a Mach-Zehnder modulator. The modulator is based on seamless integration of a high-contrast waveguide based on lithium niobate-a popular modulator material-with compact, low-loss silicon circuitry. The present device demonstrates a modulation efficiency as high as 1.2 Vcm and a low insertion loss of 3.3 dB. The 3dB electro-optic bandwidth is approximately 17.5 GHz. The optical eye diagrams, operating at 32 Gbit/s and 40 Gbit/s, with measured dynamic extinction ratios at 8 dB and 6.6 dB respectively. The present device avoids absorption loss and nonlinearity in conventional silicon modulators and demonstrates highest modulation efficiency in LN modulators, showing potential in future optical interconnects.

physics.app-ph↗

arXiv2019

Person Re-identification Using Visual Attention

Despite recent attempts for solving the person re-identification problem, it remains a challenging task since a person's appearance can vary significantly when large variations in view angle, human pose, and illumination are involved. In this paper, we propose a novel approach based on using a gradient-based attention mechanism in deep convolution neural network for solving the person re-identification problem. Our model learns to focus selectively on parts of the input image for which the networks' output is most sensitive to and processes them with high resolution while perceiving the surrounding image in low resolution. Extensive comparative evaluations demonstrate that the proposed method outperforms state-of-the-art approaches on the challenging CUHK01, CUHK03, and Market 1501 datasets.

cs.CV↗

arXiv2019

Lending Orientation to Neural Networks for Cross-view Geo-localization

This paper studies image-based geo-localization (IBL) problem using ground-to-aerial cross-view matching. The goal is to predict the spatial location of a ground-level query image by matching it to a large geotagged aerial image database (e.g., satellite imagery). This is a challenging task due to the drastic differences in their viewpoints and visual appearances. Existing deep learning methods for this problem have been focused on maximizing feature similarity between spatially close-by image pairs, while minimizing other images pairs which are far apart. They do so by deep feature embedding based on visual appearance in those ground-and-aerial images. However, in everyday life, humans commonly use {\em orientation} information as an important cue for the task of spatial localization. Inspired by this insight, this paper proposes a novel method which endows deep neural networks with the `commonsense' of orientation. Given a ground-level spherical panoramic image as query input (and a large georeferenced satellite image database), we design a Siamese network which explicitly encodes the orientation (i.e., spherical directions) of each pixel of the images. Our method significantly boosts the discriminative power of the learned deep features, leading to a much higher recall and precision outperforming all previous methods. Our network is also more compact using only 1/5th number of parameters than a previously best-performing network. To evaluate the generalization of our method, we also created a large-scale cross-view localization benchmark containing 100K geotagged ground-aerial pairs covering a city. Our codes and datasets are available at \url{https://github.com/Liumouliu/OriCNN}.

cs.CV↗

arXiv2019

FurcaNet: An end-to-end deep gated convolutional, long short-term memory, deep neural networks for single channel speech separation

Deep gated convolutional networks have been proved to be very effective in single channel speech separation. However current state-of-the-art framework often considers training the gated convolutional networks in time-frequency (TF) domain. Such an approach will result in limited perceptual score, such as signal-to-distortion ratio (SDR) upper bound of separated utterances and also fail to exploit an end-to-end framework. In this paper we present an integrated simple and effective end-to-end approach to monaural speech separation, which consists of deep gated convolutional neural networks (GCNN) that takes the mixed utterance of two speakers and maps it to two separated utterances, where each utterance contains only one speaker's voice. In addition long short-term memory (LSTM) is employed for long term temporal modeling. For the objective, we propose to train the network by directly optimizing utterance level SDR in a permutation invariant training (PIT) style. Our experiments on the public WSJ0-2mix data corpus demonstrate that this new scheme can produce more discriminative separated utterances and leading to performance improvement on the speaker separation task.

cs.SD↗

arXiv2019

Approximate Newton-based statistical inference using only stochastic gradients

We present a novel statistical inference framework for convex empirical risk minimization, using approximate stochastic Newton steps. The proposed algorithm is based on the notion of finite differences and allows the approximation of a Hessian-vector product from first-order information. In theory, our method efficiently computes the statistical error covariance in $M$-estimation, both for unregularized convex learning problems and high-dimensional LASSO regression, without using exact second order information, or resampling the entire data set. We also present a stochastic gradient sampling scheme for statistical inference in non-i.i.d. time series analysis, where we sample contiguous blocks of indices. In practice, we demonstrate the effectiveness of our framework on large-scale machine learning problems, that go even beyond convexity: as a highlight, our work can be used to detect certain adversarial attacks on neural networks.

cs.LG↗

arXiv2019

Is CQT more suitable for monaural speech separation than STFT? an empirical study

Short-time Fourier transform (STFT) is used as the front end of many popular successful monaural speech separation methods, such as deep clustering (DPCL), permutation invariant training (PIT) and their various variants. Since the frequency component of STFT is linear, while the frequency distribution of human auditory system is nonlinear. In this work we propose and give an empirical study to use an alternative front end called constant Q transform (CQT) instead of STFT to achieve a better simulation of the frequency resolving power of the human auditory system. The upper bound in signal-to-distortion (SDR) of ideal speech separation based on CQT's ideal ration mask (IRM) is higher than that based on STFT. In the same experimental setting on WSJ0-2mix corpus, we examined the performance of CQT under different backends, including the original DPCL, utterance level PIT, and some of their variants. It is found that all CQT-based methods are better than STFT-based methods, and achieved on average 0.4dB better performance than STFT based method in SDR improvements.

cs.SD↗

arXiv2019

Micro-macro decomposition based asymptotic-preserving numerical schemes and numerical moments conservation for collisional nonlinear kinetic equations

In this paper, we first extend the micro-macro decomposition method for multiscale kinetic equations from the BGK model to general collisional kinetic equations, including the Boltzmann and the Fokker-Planck Landau equations. The main idea is to use a relation between the (numerically stiff) linearized collision operator with the nonlinear quadratic ones, the later's stiffness can be overcome using the BGK penalization method of Filbet and Jin for the Boltzmann, or the linear Fokker-Planck penalization method of Jin and Yan for the Fokker-Planck Landau equations. Such a scheme allows the computation of multiscale collisional kinetic equations efficiently in all regimes, including the fluid regime in which the fluid dynamic behavior can be correctly computed even without resolving the small Knudsen number. A distinguished feature of these schemes is that although they contain implicit terms, they can be implemented explicitly. These schemes preserve the moments (mass, momentum and energy) exactly thanks to the use of the macroscopic system which is naturally in a conservative form. We further utilize this conservation property for more general kinetic systems, using the Vlasov-Ampére and Vlasov-Ampére-Boltzmann systems as examples. The main idea is to evolve both the kinetic equation for the probability density distribution and the moment system, the later naturally induces a scheme that conserves exactly the moments numerically if they are physically conserved.

math.NA↗

arXiv2019

Spectral Convergence of the Stochastic Galerkin Approximation to the Boltzmann Equation with Multiple Scales and Large Random Perturbation in the Collision Kernel

In [L. Liu and S. Jin, Multiscale Model. Simult., 16, 1085-1114, 2018], spectral convergence and long-time decay of the numerical solution towards the global equilibrium of the stochastic Galerkin approximation for the Boltzmann equation with random inputs in the initial data and collision kernel for hard potentials and Maxwellian molecules under Grad's angular cutoff were established using the hypocoercive properties of the collisional kinetic model. One assumption for the random perturbation of the collision kernel is that the perturbation is in the order of the Knudsen number, which can be very small in the fluid dynamical regime. In this article, we remove this smallness assumption, and establish the same results but now for random perturbations of the collision kernel that can be of order one. The new analysis relies on the establishment of a spectral gap for the numerical collision operator.

math.AP↗

arXiv2018

High-Performance Hybrid Silicon and Lithium Niobate Mach-Zehnder Modulators for 100 Gbit/s and Beyond

Optical modulators are at the heart of optical communication links. Ideally, they should feature low insertion loss, low drive voltage, large modulation bandwidth, high linearity, compact footprint and low manufacturing cost. Unfortunately, these criteria have only been achieved on separate occasions.Based on a Silicon and Lithium Niobate hybrid integration platform, we demonstrate Mach-Zehnder modulators that simultaneously fulfill these criteria. The presented device exhibits an insertion loss of 2.5 dB, voltage-length product of 2.2 Vcm, high linearity, electro-optic bandwidth of at least 70 GHz and modulation rates up to 112 Gbit/s. The high-performance modulator is realized by seamless integration of high-contrast waveguide based on Lithium Niobate - the most mature modulator material - with compact, low-loss silicon circuits. The hybrid platform demonstrated here allows for the combination of 'best-in-breed' active and passive components, opening up new avenues for enabling future high-speed, energy efficient and cost-effective optical communication networks.

physics.app-ph↗