Searcharxiv⌕ Search

arXiv subjects

Hairong Liu

Publications and source records attributed to Hairong Liu.

29 records · Page 2Linked to original sources

Critical points of solutions to a kind of linear elliptic equations in multiply connected domains

In this paper, we mainly study the critical points and critical zero points of solutions $u$ to a kind of linear elliptic equations with nonhomogeneous Dirichlet boundary conditions in a multiply connected domain $Ω$ in $\mathbb{R}^2$. Based on the fine analysis about the distributions of connected components of the super-level sets $\{x\in Ω: u(x)>t\}$ and sub-level sets $\{x\in Ω: u(x)<t\}$ for some $t$, we obtain the geometric structure of interior critical point sets of $u$. Precisely, let $Ω$ be a multiply connected domain with the interior boundary $γ_I$ and the external boundary $γ_E$, where $u|_{γ_I}=ψ_1(x),~u|_{γ_E}=ψ_2(x)$. When $ψ_1(x)$ and $ψ_2(x)$ have $N_1$ and $N_2$ local maximal points on $γ_I$ and $γ_E$ respectively, we deduce that $\sum_{i = 1}^k {m_i}\leq N_1+ N_2$, where $m_1,\cdots,m_k$ are the respective multiplicities of interior critical points $x_1,\cdots,x_k$ of $u$. In addition, when $\min_{γ_E}ψ_2(x)\geq \max_{γ_I}ψ_1(x)$ and $u$ has only $N_1$ and $N_2$ equal local maxima relative to $\overlineΩ$ on $γ_I$ and $γ_E$ respectively, we develop a new method to show that one of the following three results holds $\sum_{i = 1}^k {m_i}=N_1+N_2$ or $\sum_{i = 1}^k {m_i}+1=N_1+N_2$ or $\sum_{i = 1}^k {m_i}+2=N_1+N_2$. Moreover, we investigate the geometric structure of interior critical zero points of $u.$ We obtain that the sum of multiplicities of the interior critical zero points of $u$ is less than or equal to the half of the number of its isolated zero points on the boundaries.

math.AP↗

Uniqueness of critical points of solutions to the mean curvature equation with Neumann and Robin boundary conditions

In this paper, we investigate the critical points of solutions to the prescribed constant mean curvature equation with Neumann and Robin boundary conditions respectively in a bounded smooth convex domain $Ω$ of $\mathbb{R}^{n}(n\geq2)$. Firstly, we show the non-degeneracy and uniqueness of the critical points of solutions in a planar domain by using the local Chen & Huang's comparison technique and the geometric properties of approximate surfaces at the non-degenerate critical points. Secondly, we deduce the uniqueness and non-degeneracy of the critical points of solutions in a rotationally symmetric domain of $\mathbb{R}^{n}(n\geq3)$ by the projection of higher dimensional space onto two dimensional plane.

math.AP↗

Resource-Efficient Neural Architect

Neural Architecture Search (NAS) is a laborious process. Prior work on automated NAS targets mainly on improving accuracy, but lacks consideration of computational resource use. We propose the Resource-Efficient Neural Architect (RENA), an efficient resource-constrained NAS using reinforcement learning with network embedding. RENA uses a policy network to process the network embeddings to generate new configurations. We demonstrate RENA on image recognition and keyword spotting (KWS) problems. RENA can find novel architectures that achieve high performance even with tight resource constraints. For CIFAR10, it achieves 2.95% test error when compute intensity is greater than 100 FLOPs/byte, and 3.87% test error when model size is less than 3M parameters. For Google Speech Commands Dataset, RENA achieves the state-of-the-art accuracy without resource constraints, and it outperforms the optimized architectures with tight resource constraints.

cs.NE↗

Critical points of solutions to a quasilinear elliptic equation with nonhomogeneous Dirichlet boundary conditions

In this paper, we mainly investigate the critical points associated to solutions $u$ of a quasilinear elliptic equation with nonhomogeneous Dirichlet boundary conditions in a connected domain $Ω$ in $\mathbb{R}^2$. Based on the fine analysis about the distribution of connected components of a super-level set $\{x\in Ω: u(x)>t\}$ for any $\mathop {\min}_{\partialΩ}u(x)<t<\mathop {\max}_{\partialΩ}u(x)$, we obtain the geometric structure of interior critical points of $u$. Precisely, when $Ω$ is simply connected, we develop a new method to prove $Σ_{i = 1}^k {m_i}+1=N$, where $m_1,\cdots,m_k$ are the respective multiplicities of interior critical points $x_1,\cdots,x_k$ of $u$ and $N$ is the number of global maximal points of $u$ on $\partialΩ$. When $Ω$ is an annular domain with the interior boundary $γ_I$ and the external boundary $γ_E$, where $u|_{γ_I}=H,~u|_{γ_E}=ψ(x)$ and $ψ(x)$ has $N$ local (global) maximal points on $γ_E$. For the case $ψ(x)\geq H$ or $ψ(x)\leq H$ or $\mathop {\min}\limits_{γ_E}ψ(x)<H<\mathop {\max}\limits_{γ_E}ψ(x)$, we show that $Σ_{i = 1}^k {m_i} \le N$ (either $Σ_{i = 1}^k {m_i}=N$ or $Σ_{i = 1}^k {m_i}+1=N$).

math.AP↗

Critical points of solutions for mean curvature equation in strictly convex and nonconvex domains

In this paper, we mainly investigate the set of critical points associated to solutions of mean curvature equation with zero Dirichlet boundary condition in a strictly convex domain and a nonconvex domain respectively. Firstly, we deduce that mean curvature equation has exactly one nondegenerate critical point in a smooth, bounded and strictly convex domain of $\mathbb{R}^{n}(n\geq2)$. Secondly, we study the geometric structure about the critical set $K$ of solutions $u$ for the constant mean curvature equation in a concentric (respectively an eccentric) spherical annulus domain of $\mathbb{R}^{n}(n\geq3)$, and deduce that $K$ exists (respectively does not exist) a rotationally symmetric critical closed surface $S$. In fact, in an eccentric spherical annulus domain, $K$ is made up of finitely many isolated critical points ($p_1,p_2,\cdots,p_l$) on an axis and finitely many rotationally symmetric critical Jordan curves ($C_1,C_2,\cdots,C_k$) with respect to an axis.

math.AP↗

Gram-CTC: Automatic Unit Selection and Target Decomposition for Sequence Labelling

Most existing sequence labelling models rely on a fixed decomposition of a target sequence into a sequence of basic units. These methods suffer from two major drawbacks: 1) the set of basic units is fixed, such as the set of words, characters or phonemes in speech recognition, and 2) the decomposition of target sequences is fixed. These drawbacks usually result in sub-optimal performance of modeling sequences. In this pa- per, we extend the popular CTC loss criterion to alleviate these limitations, and propose a new loss function called Gram-CTC. While preserving the advantages of CTC, Gram-CTC automatically learns the best set of basic units (grams), as well as the most suitable decomposition of tar- get sequences. Unlike CTC, Gram-CTC allows the model to output variable number of characters at each time step, which enables the model to capture longer term dependency and improves the computational efficiency. We demonstrate that the proposed Gram-CTC improves CTC in terms of both performance and efficiency on the large vocabulary speech recognition task at multiple scales of data, and that with Gram-CTC we can outperform the state-of-the-art on a standard speech benchmark.

cs.CL↗

Exploring Neural Transducers for End-to-End Speech Recognition

In this work, we perform an empirical comparison among the CTC, RNN-Transducer, and attention-based Seq2Seq models for end-to-end speech recognition. We show that, without any language model, Seq2Seq and RNN-Transducer models both outperform the best reported CTC models with a language model, on the popular Hub5'00 benchmark. On our internal diverse dataset, these trends continue - RNNTransducer models rescored with a language model after beam search outperform our best CTC models. These results simplify the speech recognition pipeline so that decoding can now be expressed purely as neural network operations. We also study how the choice of encoder architecture affects the performance of the three models - when all encoder layers are forward only, and when encoders downsample the input representation aggressively.

cs.CL↗

Reducing Bias in Production Speech Models

Replacing hand-engineered pipelines with end-to-end deep learning systems has enabled strong results in applications like speech and object recognition. However, the causality and latency constraints of production systems put end-to-end speech models back into the underfitting regime and expose biases in the model that we show cannot be overcome by "scaling up", i.e., training bigger models on more data. In this work we systematically identify and address sources of bias, reducing error rates by up to 20% while remaining practical for deployment. We achieve this by utilizing improved neural architectures for streaming inference, solving optimization issues, and employing strategies that increase audio and label modelling versatility.

cs.CL↗

Active Learning for Speech Recognition: the Power of Gradients

In training speech recognition systems, labeling audio clips can be expensive, and not all data is equally valuable. Active learning aims to label only the most informative samples to reduce cost. For speech recognition, confidence scores and other likelihood-based active learning methods have been shown to be effective. Gradient-based active learning methods, however, are still not well-understood. This work investigates the Expected Gradient Length (EGL) approach in active learning for end-to-end speech recognition. We justify EGL from a variance reduction perspective, and observe that EGL's measure of informativeness picks novel samples uncorrelated with confidence scores. Experimentally, we show that EGL can reduce word errors by 11\%, or alternatively, reduce the number of samples to label by 50\%, when compared to random sampling.

cs.CL↗

Revealing Cluster Structure of Graph by Path Following Replicator Dynamic

In this paper, we propose a path following replicator dynamic, and investigate its potentials in uncovering the underlying cluster structure of a graph. The proposed dynamic is a generalization of the discrete replicator dynamic. The replicator dynamic has been successfully used to extract dense clusters of graphs; however, it is often sensitive to the degree distribution of a graph, and usually biased by vertices with large degrees, thus may fail to detect the densest cluster. To overcome this problem, we introduce a dynamic parameter, called path parameter, into the evolution process. The path parameter can be interpreted as the maximal possible probability of a current cluster containing a vertex, and it monotonically increases as evolution process proceeds. By limiting the maximal probability, the phenomenon of some vertices dominating the early stage of evolution process is suppressed, thus making evolution process more robust. To solve the optimization problem with a fixed path parameter, we propose an efficient fixed point algorithm. The time complexity of the path following replicator dynamic is only linear in the number of edges of a graph, thus it can analyze graphs with millions of vertices and tens of millions of edges on a common PC in a few minutes. Besides, it can be naturally generalized to hypergraph and graph with edges of different orders. We apply it to four important problems: maximum clique problem, densest k-subgraph problem, structure fitting, and discovery of high-density regions. The extensive experimental results clearly demonstrate its advantages, in terms of robustness, scalability and flexility.

cs.LG↗

A mean value formula of sub-p-Laplace parabolic equations on the Heisenberg group

We derive two equivalent definitions of the viscosity solutions to the homogeneous sub-p- Laplace parabolic equations on the Heisenberg group, and characterize the viscosity solutions in terms of an asymptotic mean value formula. Moreover, we construct an example to show that these formulae do not hold in non-asymptotic sense.

math.AP↗