arXiv subjects

Liang Liu

Publications and source records attributed to Liang Liu.

At least 217 records · Page 12Linked to original sources

arXiv2020

HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation

Self-supervised learning shows great potential in monoculardepth estimation, using image sequences as the only source ofsupervision. Although people try to use the high-resolutionimage for depth estimation, the accuracy of prediction hasnot been significantly improved. In this work, we find thecore reason comes from the inaccurate depth estimation inlarge gradient regions, making the bilinear interpolation er-ror gradually disappear as the resolution increases. To obtainmore accurate depth estimation in large gradient regions, itis necessary to obtain high-resolution features with spatialand semantic information. Therefore, we present an improvedDepthNet, HR-Depth, with two effective strategies: (1) re-design the skip-connection in DepthNet to get better high-resolution features and (2) propose feature fusion Squeeze-and-Excitation(fSE) module to fuse feature more efficiently.Using Resnet-18 as the encoder, HR-Depth surpasses all pre-vious state-of-the-art(SoTA) methods with the least param-eters at both high and low resolution. Moreover, previousstate-of-the-art methods are based on fairly complex and deepnetworks with a mass of parameters which limits their realapplications. Thus we also construct a lightweight networkwhich uses MobileNetV3 as encoder. Experiments show thatthe lightweight network can perform on par with many largemodels like Monodepth2 at high-resolution with only20%parameters. All codes and models will be available at https://github.com/shawLyu/HR-Depth.

cs.CV↗

arXiv2020

Distributed and Scalable Uplink Processing for LIS: Algorithm, Architecture, and Design Trade-offs

The Large Intelligent Surface (LIS) is a promising technology in the areas of wireless communication, remote sensing and positioning. It consists of a continuous radiating surface located in the proximity of the users, with the capability to communicate by transmission and reception (replacing base stations). Despite of its potential, there are numerous challenges from implementation point of view, being the interconnection data-rate, computational complexity, and storage the most relevant ones. In order to address those challenges, hierarchical architectures with distributed processing techniques are envisioned to to be relevant for this task, while ensuring scalability. In this work we perform algorithm-architecture codesign to propose two distributed interference cancellation algorithms, and a tree-based interconnection topology for uplink processing. We also analyze the performance, hardware requirements, and architecture tradeoffs for a discrete LIS, in order to provide concrete case studies and guidelines for efficient implementation of LIS systems.

eess.SP↗

arXiv2020

Millimeter-Wave Massive MIMO Testbed with Hybrid Beamforming

Massive multiple-input multiple-out (MIMO) technology is vital in millimeter-wave (mmWave) bands to obtain large array gains. However, there are practical challenges, such as high hardware cost and power consumption in such systems. A promising solution to these problems is to adopt a hybrid beamforming architecture. This architecture has a much lower number of transceiver (TRx) chains than the total antenna number, resulting in cost- and energy-efficient systems. In this paper, we present a real-time mmWave (28 GHz) massive MIMO testbed with hybrid beamforming. This testbed has a 64-antenna/16-TRx unit for beam-selection, which can be expanded to larger array sizes in a modular way. For testing everything from baseband processing algorithms to scheduling and beam-selection in real propagation environments, we extend the capability of an existing 100-antenna/100-TRx massive MIMO testbed (below 6 GHz), built upon software-defined radio technology, to a flexible mmWave massive MIMO system.

eess.SP↗

arXiv2020

Learning by Analogy: Reliable Supervision from Transformations for Unsupervised Optical Flow Estimation

Unsupervised learning of optical flow, which leverages the supervision from view synthesis, has emerged as a promising alternative to supervised methods. However, the objective of unsupervised learning is likely to be unreliable in challenging scenes. In this work, we present a framework to use more reliable supervision from transformations. It simply twists the general unsupervised learning pipeline by running another forward pass with transformed data from augmentation, along with using transformed predictions of original data as the self-supervision signal. Besides, we further introduce a lightweight network with multiple frames by a highly-shared flow decoder. Our method consistently gets a leap of performance on several benchmarks with the best accuracy among deep unsupervised methods. Also, our method achieves competitive results to recent fully supervised methods while with much fewer parameters.

cs.CV↗

arXiv2020

Model-independently calibrating the luminosity correlations of gamma-ray bursts using deep learning

Gamma-ray bursts (GRBs) detected at high redshift can be used to trace the Hubble diagram of the Universe. However, the distance calibration of GRBs is not as easily as that of type Ia supernovae (SNe Ia). For the calibrating method based on the empirical luminosity correlations, there is an underlying assumption that the correlations should be universal over the whole redshift range. In this paper, we investigate the possible redshift dependence of six luminosity correlations with a completely model-independent deep learning method. We construct a network combining the Recurrent Neural Networks (RNN) and the Bayesian Neural Networks (BNN), where RNN is used to reconstruct the distance-redshift relation by training the network with the Pantheon compilation, and BNN is used to calculate the uncertainty of the reconstruction. Using the reconstructed distance-redshift relation of Pantheon, we test the redshift dependence of six luminosity correlations by dividing the full GRB sample into two subsamples (low-$z$ and high-$z$ subsamples), and find that only the $E_p-E_γ$ relation has no evidence for redshift dependence. We use the $E_p-E_γ$ relation to calibrate GRBs, and the calibrated GRBs give tight constraint on the flat $Λ$CDM model, with the best-fitting parameter $Ω_{\rm M}$=0.307$^{+0.065}_{-0.073}$.

astro-ph.CO↗

arXiv2020

A Two-Stage Radar Sensing Approach based on MIMO-OFDM Technology

Recently, integrating the communication and sensing functions into a common network has attracted a great amount of attention. This paper considers the advanced signal processing techniques for enabling the radar to sense the environment via the communication signals. Since the technologies of orthogonal frequency division multiplexing (OFDM) and multiple-input multiple-output (MIMO) are widely used in the legacy cellular systems, this paper proposes a two-stage signal processing approach for radar sensing in an MIMO-OFDM system, where the scattered channels caused by various targets are estimated in the first stage, and the location information of the targets is then extracted from their scattered channels in the second stage. Specifically, based on the observations that radar sensing is similar to multi-path communication in the sense that different targets scatter the signal sent by the radar transmitter to the radar receiver with various delay, and that the number of scatters is limited, we show that the OFDM-based channel training approach together with the compressed sensing technique can be utilized to estimate the scattered channels efficiently in Stage I. Moreover, to tackle the challenge arising from range resolution for sensing the location of closely spaced targets, we show that the MIMO radar technique can be leveraged in Stage II such that the radar has sufficient spatial samples to even detect the targets in close proximity based on their scattered channels. Last, numerical examples are provided to show the effectiveness of our proposed sensing approach which merely relies on the existing MIMO-OFDM communication techniques.

eess.SP↗

arXiv2020

QPS-r: A Cost-Effective Crossbar Scheduling Algorithm and Its Stability and Delay Analysis

In an input-queued switch, a crossbar schedule, or a matching between the input ports and the output ports needs to be computed in each switching cycle, or time slot. Designing switching algorithms with very low computational complexity, that lead to high throughput and small delay is a challenging problem. There appears to be a fundamental tradeoff between the computational complexity of the switching algorithm and the resultants throughput and delay. Parallel maximal matching algorithms (adapted for switching) appear to have stricken a sweet spot in this tradeoff, and prior work has shown the following performance guarantees. Using maximal matchings in every time slot results in at least 50% switch throughput and order-optimal (i.e., independent of the switch size N) average delay bounds for various traffic arrival processes. On the other hand, their computational complexity can be as low as $O(log^2N)$ per port/processor, which is much lower than those of the algorithms such as maximum weighted matching which ensures better throughput performance. In this work, we propose QPS-r, a parallel iterative switching algorithm that has the lowest possible computational complexity: O(1) per port. Using Lyapunov stability analysis, we show that the throughput and delay performance is identical to that of maximal matching algorithm. Although QPS-r builds upon an existing technique called Queue-Proportional Sampling (QPS), in this paper, we provide analytical guarantees on its throughput and delay under i.i.d. traffic as well as a Markovian traffic model which can model many realistic traffic patterns. We also demonstrate that QPS-3 (running 3 iterations) has comparable empirical throughput and delay performances as iSLIP (running $log_2 N$ iterations), a refined and optimized representative maximal matching algorithm adapted for switching.

cs.PF↗

arXiv2020

Space- and Computationally-Efficient Set Reconciliation via Parity Bitmap Sketch (PBS)

Set reconciliation is a fundamental algorithmic problem that arises in many networking, system, and database applications. In this problem, two large sets A and B of objects (bitcoins, files, records, etc.) are stored respectively at two different network-connected hosts, which we name Alice and Bob respectively. Alice and Bob communicate with each other to learn $AΔB$, the difference between A and B, and as a result the reconciled set $A\bigcup B$. Current set reconciliation schemes are based on either Invertible Bloom Filters (IBF) or Error-Correction Codes (ECC). The former has a low computational complexity of O(d), where d is the cardinality of $AΔB$, but has a high communication overhead that is several times larger than the theoretical minimum. The latter has a low communication overhead close to the theoretical minimum, but has a much higher computational complexity of $O(d^2)$. In this work, we propose Parity Bitmap Sketch (PBS), an ECC- based set reconciliation scheme that gets the better of both worlds: PBS has both a low computational complexity of O(d) just like IBF-based solutions and a low communication overhead of roughly twice the theoretical minimum. A separate contribution of this work is a novel rigorous analytical framework that can be used for the precise calculation of various performance metrics and for the near-optimal parameter tuning of PBS.

cs.DB↗

arXiv2020

Weighing Counts: Sequential Crowd Counting by Reinforcement Learning

We formulate counting as a sequential decision problem and present a novel crowd counting model solvable by deep reinforcement learning. In contrast to existing counting models that directly output count values, we divide one-step estimation into a sequence of much easier and more tractable sub-decision problems. Such sequential decision nature corresponds exactly to a physical process in reality scale weighing. Inspired by scale weighing, we propose a novel 'counting scale' termed LibraNet where the count value is analogized by weight. By virtually placing a crowd image on one side of a scale, LibraNet (agent) sequentially learns to place appropriate weights on the other side to match the crowd count. At each step, LibraNet chooses one weight (action) from the weight box (the pre-defined action pool) according to the current crowd image features and weights placed on the scale pan (state). LibraNet is required to learn to balance the scale according to the feedback of the needle (Q values). We show that LibraNet exactly implements scale weighing by visualizing the decision process how LibraNet chooses actions. Extensive experiments demonstrate the effectiveness of our design choices and report state-of-the-art results on a few crowd counting benchmarks. We also demonstrate good cross-dataset generalization of LibraNet. Code and models are made available at: https://git.io/libranet

cs.CV↗

arXiv2020

Magnetic asymmetry induced anomalous spin-orbit torque in IrMn

We demonstrate an anomalous spin-orbit torque induced by the broken magnetic symmetry in the antiferromagnet IrMn. We study the magnetic structure of three phases of IrMn thin films using neutron diffraction technique. The magnetic mirror symmetry M' is broken laterally in both L10-IrMn and L12-IrMn3 but not γ-IrMn3. We observe an out-of-plane damping-like spin-orbit torque in both L10-IrMn/permalloy and L12-IrMn3/permalloy bilayers but not in γ-IrMn3/permalloy. This is consistent with both the symmetry analysis on the effects of a broken M' on spin-orbit torque and the theoretical predictions of the spin Hall effect and the Rashba-Edelstein effect. In addition, the measured spin-orbit torque efficiencies are 0.61+-0.01, 1.01+-0.03 and 0.80+-0.01 for the L10, L12 and γ phases, respectively. Our work highlights the critical roles of the magnetic asymmetry in spin-orbit torque generation.

cond-mat.mtrl-sci↗

arXiv2020

The contact binary V344 Lacertae: is it a triple system?

The VRI passbands light curves of V344 Lac were presented and analyzed by using the latest version of the W-D code. The observed spectrum reveals that V344 Lac is not an A3 type but would be a later F type star according to the yielded temperature. The results of solution show that V344 Lac is an A-subtype contact binary, with a mediate photometric mass ratio of 0.387(0.003) and a mediate contact factor of 44.6(3.0)%. Based on the parallax given by Gaia, the parameters of the components are estimated as: M1 = 1.16 Ms, M2 = 0.45 Ms, R1 = 1.31 Rs, R2 = 0.88 Rs, L1 = 2.512 Ls, L2 = 1.057 Ls. The period investigation indicates that V344 Lac may have an eccentric orbital oscillation, with P3 = 12.4(0.5) yr, A3 = 0.0020(0.0002) d, and e = 0.38(0.16). Analysis shows such oscillation would be caused by a magnetic activity which can be explained by the Applegate mechanism. Meanwhile, according to the value of l3 and the estimated physical parameters of V344 Lac, the mass of the third companion may be 0.79 Ms. This third body could be a wide company.

astro-ph.SR↗

arXiv2020

Perpendicular magnetic anisotropy and Dzyaloshinskii-Moriya interaction at an oxide/ferromagnetic metal interface

We report on the study of both perpendicular magnetic anisotropy (PMA) and Dzyaloshinskii-Moriya interaction (DMI) at an oxide/ferromagnetic metal (FM) interface, i.e. BaTiO3 (BTO)/CoFeB. Thanks to the functional properties of the BTO film and the capability to precisely control its growth, we are able to distinguish the dominant role of the oxide termination (TiO2 vs BaO), from the moderate effect of ferroelectric polarization in the BTO film, on the PMA and DMI at the oxide/FM interface. We find that the interfacial magnetic anisotropy energy of the BaO-BTO/CoFeB structure is two times larger than that of the TiO2-BTO/CoFeB, while the DMI of the TiO2-BTO/CoFeB interface is larger. We explain the observed phenomena by first-principles calculations, which ascribe them to the different electronic states around the Fermi level at the oxide/ferromagnetic metal interfaces and the different spin-flip processes. This study paves the way for further investigation of the PMA and DMI at various oxide/FM structures and thus their applications in the promising field of energy-efficient devices.

cond-mat.mtrl-sci↗

arXiv2020

Channel Estimation for Intelligent Reflecting Surface Assisted Multiuser Communications: Framework, Algorithms, and Analysis

In intelligent reflecting surface (IRS) assisted communication systems, the acquisition of channel state information (CSI) is a crucial impediment for achieving the beamforming gain of IRS because of the considerable overhead required for channel estimation. Specifically, under the current beamforming design for IRS-assisted communications, $KMN+KM$ channel coefficients should be estimated, where $K$, $N$ and $M$ denote the numbers of users, IRS reflecting elements, and antennas at the base station (BS), respectively. To accurately estimate such a large number of channel coefficients within a short time interval, we propose a novel three-phase pilot-based channel estimation framework in this paper for IRS-assisted uplink multiuser communications. Under this framework, we analytically prove that a time duration consisting of $K+N+\max(K-1,\lceil (K-1)N/M \rceil)$ pilot symbols is sufficient for the BS to perfectly recover all the $KMN+KM$ channel coefficients for the case without receiver noise at the BS. In contrast to the channel estimation for conventional uplink communications without IRS where the minimum channel estimation time is independent of the number of receive antennas at the BS, our result reveals the crucial role of massive MIMO (multiple-input multiple-output) in reducing the channel estimation time for IRS-assisted communications. Further, for the case with receiver noise, the user pilot sequences, IRS reflecting coefficients, and BS linear minimum mean-squared error (LMMSE) channel estimators are characterized in closed-form, and the corresponding estimation mean-squared error (MSE) is quantified.

cs.IT↗

arXiv2020

Focal Loss Analysis of Nerve Fiber Layer Reflectance for Glaucoma Diagnosis

Purpose: To evaluate nerve fiber layer (NFL) reflectance for glaucoma diagnosis. Methods: Participants were imaged with 4.5X4.5-mm volumetric disc scans using spectral-domain optical coherence tomography (OCT). The normalized NFL reflectance map was processed by an azimuthal filter to reduce directional reflectance bias due to variation of beam incidence angle. The peripapillary area of the map was divided into 160 superpixels. Average reflectance was the mean of superpixel reflectance. Low-reflectance superpixels were identified as those with NFL reflectance below the 5 percentile normative cutoff. Focal reflectance loss was measure by summing loss in low-reflectance superpixels. Results: Thirty-five normal, 30 pre-perimetric and 35 perimetric glaucoma participants were enrolled. Azimuthal filtering improved the repeatability of the normalized NFL reflectance, as measured by the pooled superpixel standard deviation (SD), from 0.73 to 0.57 dB (p<0.001, paired t-test) and reduced the population SD from 2.14 to 1.78 dB (p<0.001, t-test). Most glaucomatous reflectance maps showed characteristic patterns of contiguous wedge or diffuse defects. Focal NFL reflectance loss had significantly higher diagnostic sensitivity than the best NFL thickness parameter (overall, inferior, or focal loss volume): 53% v. 23% (p=0.027) in PPG eyes and 100% v. 80% (p=0.023) in PG eyes, with the specificity fixed at 99%. Conclusions: Azimuthal filtering reduces the variability of NFL reflectance measurements. Focal NFL reflectance loss has excellent glaucoma diagnostic accuracy compared to the standard NFL thickness parameters. The reflectance map may be useful for localizing NFL defects.

eess.IV↗

arXiv2020

Nearly nondestructive thermometry of labeled cold atoms and application to isotropic laser cooling

We have designed and implemented a straightforward method to deterministically measure the temperature of the selected segment of a cold atom ensemble, and we have also developed an upgrade in the form of nondestructive thermometry. The essence is to monitor the thermal expansion of the targeted cold atoms after labeling them through manipulating the internal states, and the nondestructive property relies upon the nearly lossless detection via driving a cycling transition. For cold atoms subject to isotropic laser cooling, this method has the unique capability of addressing only the atoms on the optical detection axis within the enclosure, which is exactly the part we care about in major applications such as atomic clock or quantum sensing. Furthermore, our results confirm the sub-Doppler cooling features in isotropic laser cooling, and we have investigated the relevant cooling properties. Meanwhile, we have applied the recently developed optical configuration with the cooling laser injection in the form of hollow beams, which helps to enhance the cooling performance and accumulate more cold atoms in the central regions.

physics.atom-ph↗

arXiv2020

Processing Distribution and Architecture Tradeoff for Large Intelligent Surface Implementation

The Large Intelligent Surface (LIS) concept has emerged recently as a new paradigm for wireless communication, remote sensing and positioning. It consists of a continuous radiating surface placed relatively close to the users, which is able to communicate with users by independent transmission and reception (replacing base stations). Despite of its potential, there are a lot of challenges from an implementation point of view, with the interconnection data-rate and computational complexity being the most relevant. Distributed processing techniques and hierarchical architectures are expected to play a vital role addressing this while ensuring scalability. In this paper we perform algorithm-architecture codesign and analyze the hardware requirements and architecture trade-offs for a discrete LIS to perform uplink detection. By doing this, we expect to give concrete case studies and guidelines for efficient implementation of LIS systems.

eess.SP↗

arXiv2020

From Open Set to Closed Set: Supervised Spatial Divide-and-Conquer for Object Counting

Visual counting, a task that aims to estimate the number of objects from an image/video, is an open-set problem by nature, i.e., the number of population can vary in [0, inf) in theory. However, collected data and labeled instances are limited in reality, which means that only a small closed set is observed. Existing methods typically model this task in a regression manner, while they are prone to suffer from an unseen scene with counts out of the scope of the closed set. In fact, counting has an interesting and exclusive property---spatially decomposable. A dense region can always be divided until sub-region counts are within the previously observed closed set. We therefore introduce the idea of spatial divide-and-conquer (S-DC) that transforms open-set counting into a closed-set problem. This idea is implemented by a novel Supervised Spatial Divide-and-Conquer Network (SS-DCNet). Thus, SS-DCNet can only learn from a closed set but generalize well to open-set scenarios via S-DC. SS-DCNet is also efficient. To avoid repeatedly computing sub-region convolutional features, S-DC is executed on the feature map instead of on the input image. We provide theoretical analyses as well as a controlled experiment on toy data, demonstrating why closed-set modeling makes sense. Extensive experiments show that SS-DCNet achieves the state-of-the-art performance. Code and models are available at: https://tinyurl.com/SS-DCNet.

cs.CV↗

arXiv2020

A New Accelerated Stochastic Gradient Method with Momentum

In this paper, we propose a novel accelerated stochastic gradient method with momentum, which momentum is the weighted average of previous gradients. The weights decays inverse proportionally with the iteration times. Stochastic gradient descent with momentum (Sgdm) use weights that decays exponentially with the iteration times to generate an momentum term. Using exponentially decaying weights, variants of Sgdm with well designed and complicated formats have been proposed to achieve better performance. The momentum update rules of our method is as simple as that of Sgdm. We provide theoretical convergence properties analyses for our method, which show both the exponentially decay weights and our inverse proportionally decay weights can limit the variance of the moving direction of parameters to be optimized to a region. Experimental results empirically show that our method works well with practical problems and outperforms Sgdm, and it outperforms Adam in convolutional neural networks.

cs.LG↗