arXiv subjects

Kun Wang

Publications and source records attributed to Kun Wang.

At least 415 records · Page 23Linked to original sources

arXiv2021

Pitch Preservation In Singing Voice Synthesis

Suffering from limited singing voice corpus, existing singing voice synthesis (SVS) methods that build encoder-decoder neural networks to directly generate spectrogram could lead to out-of-tune issues during the inference phase. To attenuate these issues, this paper presents a novel acoustic model with independent pitch encoder and phoneme encoder, which disentangles the phoneme and pitch information from music score to fully utilize the corpus. Specifically, according to equal temperament theory, the pitch encoder is constrained by a pitch metric loss that maps distances between adjacent input pitches into corresponding frequency multiples between the encoder outputs. For the phoneme encoder, based on the analysis that same phonemes corresponding to varying pitches can produce similar pronunciations, this encoder is followed by an adversarially trained pitch classifier to enforce the identical phonemes with different pitches mapping into the same phoneme feature space. By these means, the sparse phonemes and pitches in original input spaces can be transformed into more compact feature spaces respectively, where the same elements cluster closely and cooperate mutually to enhance synthesis quality. Then, the outputs of the two encoders are summed together to pass through the following decoder in the acoustic model. Experimental results indicate that the proposed approaches can characterize intrinsic structure between pitch inputs to obtain better pitch synthesis accuracy and achieve superior singing synthesis performance against the advanced baseline system.

cs.SD↗

arXiv2021

Heterogeneous Dual-Core Overlay Processor for Light-Weight CNNs

Light-weight convolutional neural networks (CNNs) have small complexity and are good candidates for low-power, high-throughput inference. Such networks are heterogeneous in terms of computation-to-communication (CTC) ratios and computation patterns between layers, especially for different layer types. Yet, existing AI processors either use homogeneous processing elements (PEs), resulting in low runtime PE efficiency, or run different layers on heterogeneous PEs in sequential, introducing resource redundancy. This paper proposes a heterogeneous dual-core architecture (dual-OPU), where one core is optimized for regular convolution layers and the other for depthwise convolution layers. PEs are homogeneous with each core. To make full use of dual-core parallelism, we develop a scheduling algorithm to concurrently execute layers for different input images on dual-core and balance parallel workload. Meanwhile, we automatically tune the PE number for a core and tune the input size for each PE to maximize throughput. Compared with a single-core processor with the same area for a single network, heterogeneous dual-OPU on average improves runtime PE efficiency and throughput by 11% and 31%, respectively. For a workload of multiple networks, dual-OPU improves average throughput by 11% compared with the state-of-the-art processors scaled to the same area. To the best of our knowledge, it is the first in-depth study on the heterogeneous dual-core processor for light-weight CNNs.

cs.AR↗

arXiv2021

Dense Coding with Locality Restriction for Decoder: Quantum Encoders vs. Super-Quantum Encoders

We investigate dense coding by imposing various locality restrictions to our decoder by employing the resource theory of asymmetry framework. In this task, the sender Alice and the receiver Bob share an entangled state. She encodes the classical information into it using a symmetric preserving encoder and sends the encoded state to Bob through a noiseless quantum channel. The decoder is limited to a measurement to satisfy a certain locality condition on the bipartite system composed of the receiving system and the preshared entanglement half. Our contributions are summarized as follows: First, we derive an achievable transmission rate for this task dependently of conditions of encoder and decoder. Surprisingly, we show that the obtained rate cannot be improved even when the decoder is relaxed to local measurements, two-way LOCCs, separable measurements, or partial transpose positive (PPT) measurements for the bipartite system. Moreover, depending on the class of allowed measurements with a locality condition, we relax the class of encoding operations to super-quantum encoders in the framework of general probability theory (GPT). That is, when our decoder is restricted to a separable measurement, theoretically, a positive operation is allowed as an encoding operation. Surprisingly, even under this type of super-quantum relaxation, the transmission rate cannot be improved. This fact highlights the universal validity of our analysis beyond quantum theory.

quant-ph↗

arXiv2021

Elongation of Curvature-Bounded Path

The paper is concerned with elongating the shortest curvature-bounded path between two oriented points to an expected length. The elongation of curvature-bounded paths to an expected length is fundamentally important to plan missions for nonholonomic-constrained vehicles in many practical applications, such as coordinating multiple nonholonomic-constrained vehicles to reach a destination simultaneously or performing a mission with a strict time window. In the paper, the explicit conditions for the existence of curvature-bounded paths joining two oriented points with an expected length are established by applying the properties of the reachability set of curvature-bounded paths. These existence conditions are numerically verifiable, allowing readily checking the existence of curvature-bounded paths between two prescribed oriented points with a desired length. In addition, once the existence conditions are met, elongation strategies are provided in the paper to get curvature-bounded paths with expected lengths. Finally, some examples of minimum-time path planning for multiple fixed-wing aerial vehicles to cooperatively achieve a triangle-shaped flight formation are presented, illustrating and verifying the developments of the paper.

math.OC↗

arXiv2021

X2Teeth: 3D Teeth Reconstruction from a Single Panoramic Radiograph

3D teeth reconstruction from X-ray is important for dental diagnosis and many clinical operations. However, no existing work has explored the reconstruction of teeth for a whole cavity from a single panoramic radiograph. Different from single object reconstruction from photos, this task has the unique challenge of constructing multiple objects at high resolutions. To conquer this task, we develop a novel ConvNet X2Teeth that decomposes the task into teeth localization and single-shape estimation. We also introduce a patch-based training strategy, such that X2Teeth can be end-to-end trained for optimal performance. Extensive experiments show that our method can successfully estimate the 3D structure of the cavity and reflect the details for each tooth. Moreover, X2Teeth achieves a reconstruction IoU of 0.681, which significantly outperforms the encoder-decoder method by $1.71X and the retrieval-based method by $1.52X. Our method can also be promising for other multi-anatomy 3D reconstruction tasks.

cs.CV↗

arXiv2021

Domain Adaptation for Underwater Image Enhancement

Recently, learning-based algorithms have shown impressive performance in underwater image enhancement. Most of them resort to training on synthetic data and achieve outstanding performance. However, these methods ignore the significant domain gap between the synthetic and real data (i.e., interdomain gap), and thus the models trained on synthetic data often fail to generalize well to real underwater scenarios. Furthermore, the complex and changeable underwater environment also causes a great distribution gap among the real data itself (i.e., intra-domain gap). However, almost no research focuses on this problem and thus their techniques often produce visually unpleasing artifacts and color distortions on various real images. Motivated by these observations, we propose a novel Two-phase Underwater Domain Adaptation network (TUDA) to simultaneously minimize the inter-domain and intra-domain gap. Concretely, a new dual-alignment network is designed in the first phase, including a translation part for enhancing realism of input images, followed by an enhancement part. With performing image-level and feature-level adaptation in two parts by jointly adversarial learning, the network can better build invariance across domains and thus bridge the inter-domain gap. In the second phase, we perform an easy-hard classification of real data according to the assessed quality of enhanced images, where a rank-based underwater quality assessment method is embedded. By leveraging implicit quality information learned from rankings, this method can more accurately assess the perceptual quality of enhanced images. Using pseudo labels from the easy part, an easy-hard adaptation technique is then conducted to effectively decrease the intra-domain gap between easy and hard samples.

cs.CV↗

arXiv2021

Regularizing Nighttime Weirdness: Efficient Self-supervised Monocular Depth Estimation in the Dark

Monocular depth estimation aims at predicting depth from a single image or video. Recently, self-supervised methods draw much attention since they are free of depth annotations and achieve impressive performance on several daytime benchmarks. However, they produce weird outputs in more challenging nighttime scenarios because of low visibility and varying illuminations, which bring weak textures and break brightness-consistency assumption, respectively. To address these problems, in this paper we propose a novel framework with several improvements: (1) we introduce Priors-Based Regularization to learn distribution knowledge from unpaired depth maps and prevent model from being incorrectly trained; (2) we leverage Mapping-Consistent Image Enhancement module to enhance image visibility and contrast while maintaining brightness consistency; and (3) we present Statistics-Based Mask strategy to tune the number of removed pixels within textureless regions, using dynamic statistics. Experimental results demonstrate the effectiveness of each component. Meanwhile, our framework achieves remarkable improvements and state-of-the-art results on two nighttime datasets.

cs.CV↗

arXiv2021

FPB: Feature Pyramid Branch for Person Re-Identification

High performance person Re-Identification (Re-ID) requires the model to focus on both global silhouette and local details of pedestrian. To extract such more representative features, an effective way is to exploit deep models with multiple branches. However, most multi-branch based methods implemented by duplication of part backbone structure normally lead to severe increase of computational cost. In this paper, we propose a lightweight Feature Pyramid Branch (FPB) to extract features from different layers of networks and aggregate them in a bidirectional pyramid structure. Cooperated by attention modules and our proposed cross orthogonality regularization, FPB significantly prompts the performance of backbone network by only introducing less than 1.5M extra parameters. Extensive experimental results on standard benchmark datasets demonstrate that our proposed FPB based model outperforms state-of-the-art methods with obvious margin as well as much less model complexity. FPB borrows the idea of the Feature Pyramid Network (FPN) from prevailing object detection methods. To our best knowledge, it is the first successful application of similar structure in person Re-ID tasks, which empirically proves that pyramid network as affiliated branch could be a potential structure in related feature embedding models. The source code is publicly available at https://github.com/anocodetest1/FPB.git.

cs.CV↗

arXiv2021

Sequential Weakly Labeled Multi-Activity Localization and Recognition on Wearable Sensors using Recurrent Attention Networks

With the popularity and development of the wearable devices such as smartphones, human activity recognition (HAR) based on sensors has become as a key research area in human computer interaction and ubiquitous computing. The emergence of deep learning leads to a recent shift in the research of HAR, which requires massive strictly labeled data. In comparison with video data, activity data recorded from accelerometer or gyroscope is often more difficult to interpret and segment. Recently, several attention mechanisms are proposed to handle the weakly labeled human activity data, which do not require accurate data annotation. However, these attention-based models can only handle the weakly labeled dataset whose sample includes one target activity, as a result it limits efficiency and practicality. In the paper, we propose a recurrent attention networks (RAN) to handle sequential weakly labeled multi-activity recognition and location tasks. The model can repeatedly perform steps of attention on multiple activities of one sample and each step is corresponding to the current focused activity. The effectiveness of the RAN model is validated on a collected sequential weakly labeled multi-activity dataset and the other two public datasets. The experiment results show that our RAN model can simultaneously infer multi-activity types from the coarse-grained sequential weak labels and determine specific locations of every target activity with only knowledge of which types of activities contained in the long sequence. It will greatly reduce the burden of manual labeling. The code of our work is available at https://github.com/KennCoder7/RAN.

eess.SP↗

arXiv2021

Sim2Sim Evaluation of a Novel Data-Efficient Differentiable Physics Engine for Tensegrity Robots

Learning policies in simulation is promising for reducing human effort when training robot controllers. This is especially true for soft robots that are more adaptive and safe but also more difficult to accurately model and control. The sim2real gap is the main barrier to successfully transfer policies from simulation to a real robot. System identification can be applied to reduce this gap but traditional identification methods require a lot of manual tuning. Data-driven alternatives can tune dynamical models directly from data but are often data hungry, which also incorporates human effort in collecting data. This work proposes a data-driven, end-to-end differentiable simulator focused on the exciting but challenging domain of tensegrity robots. To the best of the authors' knowledge, this is the first differentiable physics engine for tensegrity robots that supports cable, contact, and actuation modeling. The aim is to develop a reasonably simplified, data-driven simulation, which can learn approximate dynamics with limited ground truth data. The dynamics must be accurate enough to generate policies that can be transferred back to the ground-truth system. As a first step in this direction, the current work demonstrates sim2sim transfer, where the unknown physical model of MuJoCo acts as a ground truth system. Two different tensegrity robots are used for evaluation and learning of locomotion policies, a 6-bar and a 3-bar tensegrity. The results indicate that only 0.25\% of ground truth data are needed to train a policy that works on the ground truth system when the differentiable engine is used for training against training the policy directly on the ground truth system.

cs.RO↗

arXiv2021

Hot subdwarf Atmospheric parameters, Kinematics, and Origins -- Based on 1587 hot subdwarf stars observed in Gaia DR2 and LAMOST DR7

Based on the Gaia DR2 catalogue of hot subdwarf star candidates, we identified 1587 hot subdwarf stars with spectra in LAMOST DR7. We present atmospheric parameters for these stars by fitting the LAMOST spectra with {\sc Tlusty/Synspec} non-LTE synthetic spectra. Combining LAMOST radial velocities and Gaia Early Data Release 3 (EDR3) parallaxes and proper motions, we also present the Galactic space positions, velocity vectors, orbital parameters and the Galactic population memberships of the stars. With our He classification scheme, we identify four groups of He rich hot subdwarf stars in the $T_{\rm eff}-\log\,g$ and $T_{\rm eff}-\log{(n{\rm He}/n{\rm H})}$ diagrams. We find two extreme He-rich groups ($e$He-1 and $e$He-2) for stars with $\log{(n{\rm He}/n{\rm H})}\geq0$ and two intermediate He-rich groups ($i$He-1 and $i$He-2) for stars with $-1\le\log{(n{\rm He}/n{\rm H})}<0$. We also find that over half of the stars in Group $e$He-1 are thick disk stars, while over half of the stars in Group $e$He-2 correspond to thin disk stars. The disk population fractions of Group $i$He-1 are between those of Group $e$He-1 and $e$He-2. Almost all stars in Group $i$He-2 belong to the thin disk. These differences indicate that the four groups probably have very different origins. Comparisons between hot subdwarf stars in the halo and in the Galactic globular cluster $ω$ Cen show that only He-deficient stars with $-2.2\le\log{(n{\rm He}/n{\rm H})}<-1$ have similar fractions. Hot subdwarfs with $\log{(n{\rm He}/n{\rm H})}\ge 0$ in $ω$ Cen have no counterparts in the thick disk and halo populations, but they appear in the thin disk.

astro-ph.SR↗

arXiv2021

Predicting Fault Slip via Transfer Learning

Data-driven machine-learning for predicting instantaneous and future fault-slip in laboratory experiments has recently progressed markedly due to large training data sets. In Earth however, earthquake interevent times range from 10's-100's of years and geophysical data typically exist for only a portion of an earthquake cycle. Sparse data presents a serious challenge to training machine learning models. Here we describe a transfer learning approach using numerical simulations to train a convolutional encoder-decoder that predicts fault-slip behavior in laboratory experiments. The model learns a mapping between acoustic emission histories and fault-slip from numerical simulations, and generalizes to produce accurate results using laboratory data. Notably slip-predictions markedly improve using the simulation-data trained-model and training the latent space using a portion of a single laboratory earthquake-cycle. The transfer learning results elucidate the potential of using models trained on numerical simulations and fine-tuned with small geophysical data sets for potential applications to faults in Earth.

physics.geo-ph↗

arXiv2021

A polygonal scaled boundary finite element method for solving heat conduction problems

This paper presents a steady-state and transient heat conduction analysis framework using the polygonal scaled boundary finite element method (PSBFEM) with polygon/quadtree meshes. The PSBFEM is implemented with commercial finite element code Abaqus by the User Element Sub-routine (UEL) feature. The detailed implementation of the framework, defining the UEL element, and solving the stiffness/mass matrix by the eigenvalue decomposition are presented. Several benchmark problems from heat conduction are solved to validate the proposed implementation. Results show that the PSBFEM is reliable and accurate for solving heat conduction problems. Not only can the proposed implementation help engineering practitioners analyze the heat conduction problem using polygonal mesh in Abaqus, but it also provides a reference for developing the UEL to solve other problems using the scaled boundary finite element method.

math.NA↗

arXiv2021

Generic energy formalism for reciprocal quadruplets within the two-sublattice quasichemical model

Ever-increasing interests to more accurate thermodynamic predictions of phase diagrams motivate more reliable thermodynamic models to be developed. The Modified Quasichemical Model within the two-sublattice Quadruplet Approximation (MQMQA) was thus established in well response to these interests and motivations. However, the model still needs to be further improved in order to have better thermodynamic predictions of arbitrary reciprocal solutions. The present paper proposes a new and generic formalism to characterize the Gibbs energy of the ternary reciprocal quadruplet within the framework of the MQMQA. The new formalism circumvents the problem spawned from solving the singular matrix of mass equations. Resultantly, energy landscapes of reciprocal solutions can be better defined everywhere in reciprocal composition spaces. With the current improvement, the MQMQA is believed to be one of the most reliable thermodynamic models for various types of solutions with or without short-range ordering.

cond-mat.mtrl-sci↗

arXiv2021

Cascading Bandit under Differential Privacy

This paper studies \emph{differential privacy (DP)} and \emph{local differential privacy (LDP)} in cascading bandits. Under DP, we propose an algorithm which guarantees $ε$-indistinguishability and a regret of $\mathcal{O}((\frac{\log T}ε)^{1+ξ})$ for an arbitrarily small $ξ$. This is a significant improvement from the previous work of $\mathcal{O}(\frac{\log^3 T}ε)$ regret. Under ($ε$,$δ$)-LDP, we relax the $K^2$ dependence through the tradeoff between privacy budget $ε$ and error probability $δ$, and obtain a regret of $\mathcal{O}(\frac{K\log (1/δ) \log T}{ε^2})$, where $K$ is the size of the arm subset. This result holds for both Gaussian mechanism and Laplace mechanism by analyses on the composition. Our results extend to combinatorial semi-bandit. We show respective lower bounds for DP and LDP cascading bandits. Extensive experiments corroborate our theoretic findings.

cs.LG↗

arXiv2021

Conservative Contextual Combinatorial Cascading Bandit

Conservative mechanism is a desirable property in decision-making problems which balance the tradeoff between the exploration and exploitation. We propose the novel \emph{conservative contextual combinatorial cascading bandit ($C^4$-bandit)}, a cascading online learning game which incorporates the conservative mechanism. At each time step, the learning agent is given some contexts and has to recommend a list of items but not worse than the base strategy and then observes the reward by some stopping rules. We design the $C^4$-UCB algorithm to solve the problem and prove its n-step upper regret bound for two situations: known baseline reward and unknown baseline reward. The regret in both situations can be decomposed into two terms: (a) the upper bound for the general contextual combinatorial cascading bandit; and (b) a constant term for the regret from the conservative mechanism. We also improve the bound of the conservative contextual combinatorial bandit as a by-product. Experiments on synthetic data demonstrate its advantages and validate our theoretical analysis.

cs.LG↗

arXiv2021

NPE: An FPGA-based Overlay Processor for Natural Language Processing

In recent years, transformer-based models have shown state-of-the-art results for Natural Language Processing (NLP). In particular, the introduction of the BERT language model brought with it breakthroughs in tasks such as question answering and natural language inference, advancing applications that allow humans to interact naturally with embedded devices. FPGA-based overlay processors have been shown as effective solutions for edge image and video processing applications, which mostly rely on low precision linear matrix operations. In contrast, transformer-based NLP techniques employ a variety of higher precision nonlinear operations with significantly higher frequency. We present NPE, an FPGA-based overlay processor that can efficiently execute a variety of NLP models. NPE offers software-like programmability to the end user and, unlike FPGA designs that implement specialized accelerators for each nonlinear function, can be upgraded for future NLP models without requiring reconfiguration. We demonstrate that NPE can meet real-time conversational AI latency targets for the BERT language model with $4\times$ lower power than CPUs and $6\times$ lower power than GPUs. We also show NPE uses $3\times$ fewer FPGA resources relative to comparable BERT network-specific accelerators in the literature. NPE provides a cost-effective and power-efficient FPGA-based solution for Natural Language Processing at the edge.

cs.AR↗

arXiv2021

Pattern formation during nonequilibrium crystallization by classical-density-functional-based approach

Solidification pattern during nonequilibrium crystallization is among the most important microstructures in the nature and technical realms. Phase field crystal (PFC) model could simulate the pattern formation during equilibrium crystallization at atom scale, but cannot grasp the nonequilibrium ones due to the absence of proper elastic-relaxation time scale. In this work, we propose a minimal classical-density-functional-theory-based model for crystal growth in supercooled liquid. Growth front nucleation (GFN) and various nonequilibrium patterns, including the faceting growth, spherulite, dendrite and the columnar-to-equiaxed transition (CET) among others, are grasped at atom scale. It is amazing that, except for undercooling and seed spacing, seed distribution is key factor that determines the CET. Overall, two-stage growth process, i.e., the diffusion-controlled growth and the GFN-dominated growth, are identified. But, compared with the second stage, the first stage becomes too short to be noticed under the high undercooling. The distinct feature for the second stage is the dramatic increments of dislocations, which explains the amorphous nucleation precursor in the supercooled liquid. Transition time between the two stages at different undercooling are investigated. Crystal growth of BCC structure further confirms our conclusions.

cond-mat.mtrl-sci↗