arXiv subjects

Lin Li

Publications and source records attributed to Lin Li.

At least 307 records · Page 17Linked to original sources

arXiv2022

Improving Cross-lingual Speech Synthesis with Triplet Training Scheme

Recent advances in cross-lingual text-to-speech (TTS) made it possible to synthesize speech in a language foreign to a monolingual speaker. However, there is still a large gap between the pronunciation of generated cross-lingual speech and that of native speakers in terms of naturalness and intelligibility. In this paper, a triplet training scheme is proposed to enhance the cross-lingual pronunciation by allowing previously unseen content and speaker combinations to be seen during training. Proposed method introduces an extra fine-tune stage with triplet loss during training, which efficiently draws the pronunciation of the synthesized foreign speech closer to those from the native anchor speaker, while preserving the non-native speaker's timbre. Experiments are conducted based on a state-of-the-art baseline cross-lingual TTS system and its enhanced variants. All the objective and subjective evaluations show the proposed method brings significant improvement in both intelligibility and naturalness of the synthesized cross-lingual speech.

cs.SD↗

arXiv2022

What is Next when Sequential Prediction Meets Implicitly Hard Interaction?

Hard interaction learning between source sequences and their next targets is challenging, which exists in a myriad of sequential prediction tasks. During the training process, most existing methods focus on explicitly hard interactions caused by wrong responses. However, a model might conduct correct responses by capturing a subset of learnable patterns, which results in implicitly hard interactions with some unlearned patterns. As such, its generalization performance is weakened. The problem gets more serious in sequential prediction due to the interference of substantial similar candidate targets. To this end, we propose a Hardness Aware Interaction Learning framework (HAIL) that mainly consists of two base sequential learning networks and mutual exclusivity distillation (MED). The base networks are initialized differently to learn distinctive view patterns, thus gaining different training experiences. The experiences in the form of the unlikelihood of correct responses are drawn from each other by MED, which provides mutual exclusivity knowledge to figure out implicitly hard interactions. Moreover, we deduce that the unlikelihood essentially introduces additional gradients to push the pattern learning of correct responses. Our framework can be easily extended to more peer base networks. Evaluation is conducted on four datasets covering cyber and physical spaces. The experimental results demonstrate that our framework outperforms several state-of-the-art methods in terms of top-k based metrics.

cs.LG↗

arXiv2022

Josephson-Coulomb drag effect between graphene and LaAlO3/SrTiO3 interfacial superconductor

Coulomb drag refers to the phenomenon that a charge current in one electronic circuit induces a responsive current in a neighboring circuit solely through Coulomb interactions. For conventional interactions between fermionic particles such as electrons, the as-induced drag current in the passive layer is orders of magnitude weaker than the active current due to strong dielectric screening effect between the two. Here we propose a 'super' Coulomb drag effect between an active normal conductor and a passive superconductor of Josephson junction arrays, whereby the passive current can greatly exceed the active. The drag force originates from the interactions between the substantially enhanced dynamical quantum fluctuations of the superconducting phases in the passive layer and the normal electrons in the active layer. We demonstrate this effect in the devices composed of monolayer graphene and LaAlO3/SrTiO3 heterointerface, an inherently non-uniform superconductor described by Josephson junction arrays. Remarkable drag signal is observed in the superconducting transition regime of the LaAlO3/SrTiO3 interface, with its sign independent of the carrier type in the graphene layer. The estimated passive-to-active ratio can reach about 0.3 at the optimal gate voltage and the temperature dependence follows that of the typical Josephson energy between superconducting puddles. Strikingly, the ratio ought to be as large as 10^5 at zero temperature by theoretical extrapolation. From engineering perspective, our device may work as current or voltage transformers, and the drag mechanism lays the foundation for synchronizing Josephson-junction-array-based terahertz radiators.

cond-mat.mes-hall↗

arXiv2021

Observation of Aharonov-Bohm effect in PbTe nanowire networks

We report phase coherent electron transport in PbTe nanowire networks with a loop geometry. Magneto-conductance shows Aharonov-Bohm (AB) oscillations with periods of $h/e$ and $h/2e$ in flux. The amplitude of $h/2e$ oscillations is enhanced near zero magnetic field, possibly due to interference between time-reversal paths. Temperature dependence of the AB amplitudes suggests a phase coherence length $\sim$ 8 - 12 $μ$m at 50 mK. This length scale is larger than the typical geometry of PbTe-based hybrid semiconductor-superconductor nanowire devices.

cond-mat.mes-hall↗

arXiv2021

Nontrivial solution for Klein-Gordon equation coupled with Born-Infeld theory with critical growth

In this paper, we study the following system \begin{eqnarray*} \left\{ \begin{array}{ll} -Δu + V(x)u-(2ω+ϕ)ϕu=λf(u)+|u|^{4}u, \ & \text{in} \ \mathbb{R}^{3}, Δϕ+ βΔ_4ϕ= 4π(ω+ϕ) u^{2}, \ & \text{in}\ \mathbb{R}^{3},\\ \end{array} \right. \end{eqnarray*} where $f(u)$ without any growth and Ambrosetti-Rabinowitz conditions. We use cut-off function and Moser iteration to obtain the existence of nontrivial solution. Finally, as a by-product of our approaches, we get the same result for Klein-Gordon-Maxwell system.

math.AP↗

arXiv2021

Scalable Geometric Deep Learning on Molecular Graphs

Deep learning in molecular and materials sciences is limited by the lack of integration between applied science, artificial intelligence, and high-performance computing. Bottlenecks with respect to the amount of training data, the size and complexity of model architectures, and the scale of the compute infrastructure are all key factors limiting the scaling of deep learning for molecules and materials. Here, we present $\textit{LitMatter}$, a lightweight framework for scaling molecular deep learning methods. We train four graph neural network architectures on over 400 GPUs and investigate the scaling behavior of these methods. Depending on the model architecture, training time speedups up to $60\times$ are seen. Empirical neural scaling relations quantify the model-dependent scaling and enable optimal compute resource allocation and the identification of scalable molecular geometric deep learning model implementations.

cs.LG↗

arXiv2021

Reinforcement Learning based Path Exploration for Sequential Explainable Recommendation

Recent advances in path-based explainable recommendation systems have attracted increasing attention thanks to the rich information provided by knowledge graphs. Most existing explainable recommendations only utilize static knowledge graphs and ignore the dynamic user-item evolutions, leading to less convincing and inaccurate explanations. Although there are some works that realize that modelling user's temporal sequential behaviour could boost the performance and explainability of the recommender systems, most of them either only focus on modelling user's sequential interactions within a path or independently and separately of the recommendation mechanism. In this paper, we propose a novel Temporal Meta-path Guided Explainable Recommendation leveraging Reinforcement Learning (TMER-RL), which utilizes reinforcement item-item path modelling between consecutive items with attention mechanisms to sequentially model dynamic user-item evolutions on dynamic knowledge graph for explainable recommendation. Compared with existing works that use heavy recurrent neural networks to model temporal information, we propose simple but effective neural networks to capture users' historical item features and path-based context to characterize the next purchased item. Extensive evaluations of TMER on two real-world datasets show state-of-the-art performance compared against recent strong baselines.

cs.IR↗

arXiv2021

An Automated Portfolio Trading System with Feature Preprocessing and Recurrent Reinforcement Learning

We propose a novel portfolio trading system, which contains a feature preprocessing module and a trading module. The feature preprocessing module consists of various data processing operations, while in the trading part, we integrate the portfolio weight rebalance function with the trading algorithm and make the trading system fully automated and suitable for individual investors, holding a handful of stocks. The data preprocessing procedures are applied to remove the white noise in the raw data set and uncover the general pattern underlying the data set before the processed feature set is inputted into the trading algorithm. Our empirical results reveal that the proposed portfolio trading system can efficiently earn high profit and maintain a relatively low drawdown, which clearly outperforms other portfolio trading strategies.

q-fin.TR↗

arXiv2021

Selective area epitaxy of PbTe-Pb hybrid nanowires on a lattice-matched substrate

Topological quantum computing is based on braiding of Majorana zero modes encoding topological qubits. A promising candidate platform for Majorana zero modes is semiconductor-superconductor hybrid nanowires. The realization of topological qubits and braiding operations requires scalable and disorder-free nanowire networks. Selective area growth of in-plane InAs and InSb nanowires, together with shadow-wall growth of superconductor structures, have demonstrated this scalability by achieving various network structures. However, the noticeable lattice mismatch at the nanowire-substrate interface, acting as a disorder source, imposes a serious obstacle along with this roadmap. Here, combining selective area and shadow-wall growth, we demonstrate the fabrication of PbTe-Pb hybrid nanowires - another potentially promising Majorana system - on a nearly perfectly lattice-matched substrate CdTe, all done in one molecular beam epitaxy chamber. Transmission electron microscopy shows the single-crystal nature of the PbTe nanowire and its atomically sharp and clean interfaces to the CdTe substrate and the Pb overlayer, without noticeable inter-diffusion or strain. The nearly ideal interface condition, together with the strong screening of charge impurities due to the large dielectric constant of PbTe, hold promise towards a clean nanowire system to study Majorana zero modes and topological quantum computing.

cond-mat.mes-hall↗

arXiv2021

Learning Text-Image Joint Embedding for Efficient Cross-Modal Retrieval with Deep Feature Engineering

This paper introduces a two-phase deep feature engineering framework for efficient learning of semantics enhanced joint embedding, which clearly separates the deep feature engineering in data preprocessing from training the text-image joint embedding model. We use the Recipe1M dataset for the technical description and empirical validation. In preprocessing, we perform deep feature engineering by combining deep feature engineering with semantic context features derived from raw text-image input data. We leverage LSTM to identify key terms, deep NLP models from the BERT family, TextRank, or TF-IDF to produce ranking scores for key terms before generating the vector representation for each key term by using word2vec. We leverage wideResNet50 and word2vec to extract and encode the image category semantics of food images to help semantic alignment of the learned recipe and image embeddings in the joint latent space. In joint embedding learning, we perform deep feature engineering by optimizing the batch-hard triplet loss function with soft-margin and double negative sampling, taking into account also the category-based alignment loss and discriminator-based alignment loss. Extensive experiments demonstrate that our SEJE approach with deep feature engineering significantly outperforms the state-of-the-art approaches.

cs.CV↗

arXiv2021

Visual-aware Attention Dual-stream Decoder for Video Captioning

Video captioning is a challenging task that captures different visual parts and describes them in sentences, for it requires visual and linguistic coherence. The attention mechanism in the current video captioning method learns to assign weight to each frame, promoting the decoder dynamically. This may not explicitly model the correlation and the temporal coherence of the visual features extracted in the sequence frames.To generate semantically coherent sentences, we propose a new Visual-aware Attention (VA) model, which concatenates dynamic changes of temporal sequence frames with the words at the previous moment, as the input of attention mechanism to extract sequence features.In addition, the prevalent approaches widely use the teacher-forcing (TF) learning during training, where the next token is generated conditioned on the previous ground-truth tokens. The semantic information in the previously generated tokens is lost. Therefore, we design a self-forcing (SF) stream that takes the semantic information in the probability distribution of the previous token as input to enhance the current token.The Dual-stream Decoder (DD) architecture unifies the TF and SF streams, generating sentences to promote the annotated captioning for both streams.Meanwhile, with the Dual-stream Decoder utilized, the exposure bias problem is alleviated, caused by the discrepancy between the training and testing in the TF learning.The effectiveness of the proposed Visual-aware Attention Dual-stream Decoder (VADD) is demonstrated through the result of experimental studies on Microsoft video description (MSVD) corpus and MSR-Video to text (MSR-VTT) datasets.

cs.CV↗

arXiv2021

DG-Labeler and DGL-MOTS Dataset: Boost the Autonomous Driving Perception

Multi-object tracking and segmentation (MOTS) is a critical task for autonomous driving applications. The existing MOTS studies face two critical challenges: 1) the published datasets inadequately capture the real-world complexity for network training to address various driving settings; 2) the working pipeline annotation tool is under-studied in the literature to improve the quality of MOTS learning examples. In this work, we introduce the DG-Labeler and DGL-MOTS dataset to facilitate the training data annotation for the MOTS task and accordingly improve network training accuracy and efficiency. DG-Labeler uses the novel Depth-Granularity Module to depict the instance spatial relations and produce fine-grained instance masks. Annotated by DG-Labeler, our DGL-MOTS dataset exceeds the prior effort (i.e., KITTI MOTS and BDD100K) in data diversity, annotation quality, and temporal representations. Results on extensive cross-dataset evaluations indicate significant performance improvements for several state-of-the-art methods trained on our DGL-MOTS dataset. We believe our DGL-MOTS Dataset and DG-Labeler hold the valuable potential to boost the visual perception of future transportation.

cs.CV↗

arXiv2021

A Chirplet Transform-based Mode Retrieval Method for Multicomponent Signals with Crossover Instantaneous Frequencies

In nature and engineering world, the acquired signals are usually affected by multiple complicated factors and appear as multicomponent nonstationary modes. In such and many other situations, it is necessary to separate these signals into a finite number of monocomponents to represent the intrinsic modes and underlying dynamics implicated in the source signals. In this paper, we consider the mode retrieval of a multicomponent signal which has crossing instantaneous frequencies (IFs), meaning that some of the components of the signal overlap in the time-frequency domain. We use the chirplet transform (CT) to represent a multicomponent signal in the three-dimensional space of time, frequency and chirp rate and introduce a CT-based signal separation scheme (CT3S) to retrieve modes. In addition, we analyze the error bounds for IF estimation and component recovery with this scheme. We also propose a matched-filter along certain specific time-frequency lines with respect to the chirp rate to make nonstationary signals be further separated and more concentrated in the three-dimensional space of CT. Furthermore, based on the approximation of source signals with linear chirps at any local time, we propose an innovative signal reconstruction algorithm, called the group filter-matched CT3S (GFCT3S), which also takes a group of components into consideration simultaneously. GFCT3S is suitable for signals with crossing IFs. It also decreases component recovery errors when the IFs curves of different components are not crossover, but fast-varying and close to one and other. Numerical experiments on synthetic and real signals show our method is more accurate and consistent in signal separation than the empirical mode decomposition, synchrosqueezing transform, and other approaches

eess.SP↗

arXiv2021

A sufficient condition for existence of iterative roots of PM functions without the characteristic endpoints condition

For PM functions of height 1, the existence of continuous iterative roots of any order was obtained under the characteristic endpoints condition. This raises an open problem about iterative roots without this condition, called characteristic endpoints problem. This problem is solved almost completely when the number of forts is equal to or less than the order. In this paper, we study the case that the number of forts is greater than the order and give a sufficient condition for existence of continuous iterative roots of order $2$ with height 2, answering the characteristic endpoints problem partially.

math.CA↗

arXiv2021

Financial Trading with Feature Preprocessing and Recurrent Reinforcement Learning

Financial trading aims to build profitable strategies to make wise investment decisions in the financial market. It has attracted interests in the machine learning community for a long time. This paper proposes to trade financial assets automatically using feature preprocessing skills and Recurrent Reinforcement Learning (RRL) algorithm. The strategy starts from technical indicators extracted from assets' market information. Then these technical indicators are preprocessed by Principal Component Analysis (PCA) and Discrete Wavelet Transform (DWT) and eventually inputted to the RRL algorithm to do the trading. The extensive empirical evidence shows that the proposed strategy is not only effective and robust in its performance, but also can mitigate the drawbacks underlying the initial trading using RRL.

q-fin.TR↗

arXiv2021

Generate & Rank: A Multi-task Framework for Math Word Problems

Math word problem (MWP) is a challenging and critical task in natural language processing. Many recent studies formalize MWP as a generation task and have adopted sequence-to-sequence models to transform problem descriptions to mathematical expressions. However, mathematical expressions are prone to minor mistakes while the generation objective does not explicitly handle such mistakes. To address this limitation, we devise a new ranking task for MWP and propose Generate & Rank, a multi-task framework based on a generative pre-trained language model. By joint training with generation and ranking, the model learns from its own mistakes and is able to distinguish between correct and incorrect expressions. Meanwhile, we perform tree-based disturbance specially designed for MWP and an online update to boost the ranker. We demonstrate the effectiveness of our proposed method on the benchmark and the results show that our method consistently outperforms baselines in all datasets. Particularly, in the classical Math23k, our method is 7% (78.4% $\rightarrow$ 85.4%) higher than the state-of-the-art.

cs.CL↗

arXiv2021

Hyper Meta-Path Contrastive Learning for Multi-Behavior Recommendation

User purchasing prediction with multi-behavior information remains a challenging problem for current recommendation systems. Various methods have been proposed to address it via leveraging the advantages of graph neural networks (GNNs) or multi-task learning. However, most existing works do not take the complex dependencies among different behaviors of users into consideration. They utilize simple and fixed schemes, like neighborhood information aggregation or mathematical calculation of vectors, to fuse the embeddings of different user behaviors to obtain a unified embedding to represent a user's behavioral patterns which will be used in downstream recommendation tasks. To tackle the challenge, in this paper, we first propose the concept of hyper meta-path to construct hyper meta-paths or hyper meta-graphs to explicitly illustrate the dependencies among different behaviors of a user. How to obtain a unified embedding for a user from hyper meta-paths and avoid the previously mentioned limitations simultaneously is critical. Thanks to the recent success of graph contrastive learning, we leverage it to learn embeddings of user behavior patterns adaptively instead of assigning a fixed scheme to understand the dependencies among different behaviors. A new graph contrastive learning based framework is proposed by coupling with hyper meta-paths, namely HMG-CR, which consistently and significantly outperforms all baselines in extensive comparison experiments.

cs.IR↗

arXiv2021

XMUSPEECH System for VoxCeleb Speaker Recognition Challenge 2021

This paper describes the XMUSPEECH speaker recognition and diarisation systems for the VoxCeleb Speaker Recognition Challenge 2021. For track 2, we evaluate two systems including ResNet34-SE and ECAPA-TDNN. For track 4, an important part of our system is VAD module which greatly improves the performance. Our best submission on the track 4 obtained on the evaluation set DER 5.54% and JER 27.11%, while the performance on the development set is DER 2.92% and JER 20.84%.

eess.AS↗