arXiv subjects

Lei Li

Publications and source records attributed to Lei Li.

At least 685 records · Page 38Linked to original sources

arXiv2020

On the Random Batch Method for second order interacting particle systems

We investigate several important issues regarding the Random Batch Method (RBM) for second order interacting particle systems. We first show the uniform-in-time strong convergence for second order systems under suitable contraction conditions. Secondly, we propose the application of RBM for singular interaction kernels via kernel splitting strategy, and investigate numerically the application to molecular dynamics.

math.NA↗

arXiv2020

The Volctrans Machine Translation System for WMT20

This paper describes our VolcTrans system on WMT20 shared news translation task. We participated in 8 translation directions. Our basic systems are based on Transformer, with several variants (wider or deeper Transformers, dynamic convolutions). The final system includes text pre-process, data selection, synthetic data generation, advanced model ensemble, and multilingual pre-training.

cs.CL↗

arXiv2020

Deep Modulation Recognition with Multiple Receive Antennas: An End-to-end Feature Learning Approach

Modulation recognition using deep neural networks has shown promising advantages over conventional algorithms. However, most existing research focuses on single receive antenna. In this paper, two end-to-end feature learning deep architectures are introduced for modulation recognition with multiple receive antennas. The first is based on multi-view convolutional neural network by treating signals from different receive antennas as different views of a 3D object and designing the location and operation of view-pooling layer that are suitable for feature fusion of multi-antenna signals. Considering that the instantaneous SNRs could be different among receive antennas in wireless communications, we further propose weight-learning convolutional neural network which uses a weight-learning module to automatically learn the weights for feature combing of different receive antennas to perform end-to-end feature learning of multi-antenna signals. Results show that both end-to-end feature learning deep architectures outperform the existing algorithm, and the proposed weight-learning convolutional neural network achieves the best performance.

eess.SP↗

arXiv2020

On the Sentence Embeddings from Pre-trained Language Models

Pre-trained contextual representations like BERT have achieved great success in natural language processing. However, the sentence embeddings from the pre-trained language models without fine-tuning have been found to poorly capture semantic meaning of sentences. In this paper, we argue that the semantic information in the BERT embeddings is not fully exploited. We first reveal the theoretical connection between the masked language model pre-training objective and the semantic similarity task theoretically, and then analyze the BERT sentence embeddings empirically. We find that BERT always induces a non-smooth anisotropic semantic space of sentences, which harms its performance of semantic similarity. To address this issue, we propose to transform the anisotropic sentence embedding distribution to a smooth and isotropic Gaussian distribution through normalizing flows that are learned with an unsupervised objective. Experimental results show that our proposed BERT-flow method obtains significant performance gains over the state-of-the-art sentence embeddings on a variety of semantic textual similarity tasks. The code is available at https://github.com/bohanli/BERT-flow.

cs.CL↗

arXiv2020

Brain Tumor Segmentation Network Using Attention-based Fusion and Spatial Relationship Constraint

Delineating the brain tumor from magnetic resonance (MR) images is critical for the treatment of gliomas. However, automatic delineation is challenging due to the complex appearance and ambiguous outlines of tumors. Considering that multi-modal MR images can reflect different tumor biological properties, we develop a novel multi-modal tumor segmentation network (MMTSN) to robustly segment brain tumors based on multi-modal MR images. The MMTSN is composed of three sub-branches and a main branch. Specifically, the sub-branches are used to capture different tumor features from multi-modal images, while in the main branch, we design a spatial-channel fusion block (SCFB) to effectively aggregate multi-modal features. Additionally, inspired by the fact that the spatial relationship between sub-regions of tumor is relatively fixed, e.g., the enhancing tumor is always in the tumor core, we propose a spatial loss to constrain the relationship between different sub-regions of tumor. We evaluate our method on the test set of multi-modal brain tumor segmentation challenge 2020 (BraTs2020). The method achieves 0.8764, 0.8243 and 0.773 dice score for whole tumor, tumor core and enhancing tumor, respectively.

eess.IV↗

arXiv2020

Active Sentence Learning by Adversarial Uncertainty Sampling in Discrete Space

Active learning for sentence understanding aims at discovering informative unlabeled data for annotation and therefore reducing the demand for labeled data. We argue that the typical uncertainty sampling method for active learning is time-consuming and can hardly work in real-time, which may lead to ineffective sample selection. We propose adversarial uncertainty sampling in discrete space (AUSDS) to retrieve informative unlabeled samples more efficiently. AUSDS maps sentences into latent space generated by the popular pre-trained language models, and discover informative unlabeled text samples for annotation via adversarial attack. The proposed approach is extremely efficient compared with traditional uncertainty sampling with more than 10x speedup. Experimental results on five datasets show that AUSDS outperforms strong baselines on effectiveness.

cs.CL↗

arXiv2020

Volctrans Parallel Corpus Filtering System for WMT 2020

In this paper, we describe our submissions to the WMT20 shared task on parallel corpus filtering and alignment for low-resource conditions. The task requires the participants to align potential parallel sentence pairs out of the given document pairs, and score them so that low-quality pairs can be filtered. Our system, Volctrans, is made of two modules, i.e., a mining module and a scoring module. Based on the word alignment model, the mining module adopts an iterative mining strategy to extract latent parallel sentences. In the scoring module, an XLM-based scorer provides scores, followed by reranking mechanisms and ensemble. Our submissions outperform the baseline by 3.x/2.x and 2.x/2.x for km-en and ps-en on From Scratch/Fine-Tune conditions, which is the highest among all submissions.

cs.CL↗

arXiv2020

Weather conditions several hours before the strong earthquake

Physical phenomena observed before strong earthquake have been reported over centuries. Radon anomalies, electrical signals, water level changes, earthquake lights near the epicenter are recognized as pre-earthquake signals to approach earthquake prediction. Anomalous negative signals observed by ground-based atmospheric electric field instrument under fair weather open up a new way to earthquake prediction. Abnormal heat radiation before the earthquake bring fair weather around the epicenter in theory. In order to figure out the weather conditions around the epicenter before earthquakes, 213 global earthquake events with magnitude of 6 or above from 2013 to 2020 were collected. Based on our definition of fair weather, in 96.7% of the events in the statistics, the weather before the earthquake is fair. Besides, the fair state before the earthquake lasted more than 7 hours, leaving us with enough early warming time.

physics.geo-ph↗

arXiv2020

Hierarchical-environment-assisted non-Markovian and its effect on thermodynamic properties

We consider a microscopic collision model, i.e., a quantum system interacts with a hierarchical environment consisting of an auxiliary system and a reservoir. We show how the non-Markovian character of the system is influenced by the coupling strength of system-auxiliary system and auxiliary system-reservoir, initial system-environment correlations and the coherence of environment. Then we study the relation between non-Markovianity and thermodynamics properties, by studying the entropy change of system especially that from heat exchanges with memory effects, and we reveal the essence of entropy change between positive and negative values during non-Markovian evolution is due to the contribution of heat flux determined by coherence. And the information flow between the system and environment is always accompanied by energy exchange.

quant-ph↗

arXiv2020

SOLOv2: Dynamic and Fast Instance Segmentation

In this work, we aim at building a simple, direct, and fast instance segmentation framework with strong performance. We follow the principle of the SOLO method of Wang et al. "SOLO: segmenting objects by locations". Importantly, we take one step further by dynamically learning the mask head of the object segmenter such that the mask head is conditioned on the location. Specifically, the mask branch is decoupled into a mask kernel branch and mask feature branch, which are responsible for learning the convolution kernel and the convolved features respectively. Moreover, we propose Matrix NMS (non maximum suppression) to significantly reduce the inference time overhead due to NMS of masks. Our Matrix NMS performs NMS with parallel matrix operations in one shot, and yields better results. We demonstrate a simple direct instance segmentation system, outperforming a few state-of-the-art methods in both speed and accuracy. A light-weight version of SOLOv2 executes at 31.3 FPS and yields 37.1% AP. Moreover, our state-of-the-art results in object detection (from our mask byproduct) and panoptic segmentation show the potential to serve as a new strong baseline for many instance-level recognition tasks besides instance segmentation. Code is available at: https://git.io/AdelaiDet

cs.CV↗

arXiv2020

Towards Linear Time Neural Machine Translation with Capsule Networks

In this study, we first investigate a novel capsule network with dynamic routing for linear time Neural Machine Translation (NMT), referred as \textsc{CapsNMT}. \textsc{CapsNMT} uses an aggregation mechanism to map the source sentence into a matrix with pre-determined size, and then applys a deep LSTM network to decode the target sequence from the source representation. Unlike the previous work \cite{sutskever2014sequence} to store the source sentence with a passive and bottom-up way, the dynamic routing policy encodes the source sentence with an iterative process to decide the credit attribution between nodes from lower and higher layers. \textsc{CapsNMT} has two core properties: it runs in time that is linear in the length of the sequences and provides a more flexible way to select, represent and aggregates the part-whole information of the source sentence. On WMT14 English-German task and a larger WMT14 English-French task, \textsc{CapsNMT} achieves comparable results with the state-of-the-art NMT systems. To the best of our knowledge, this is the first work that capsule networks have been empirically investigated for sequence to sequence problems.

cs.CL↗

arXiv2020

Double Graph Based Reasoning for Document-level Relation Extraction

Document-level relation extraction aims to extract relations among entities within a document. Different from sentence-level relation extraction, it requires reasoning over multiple sentences across a document. In this paper, we propose Graph Aggregation-and-Inference Network (GAIN) featuring double graphs. GAIN first constructs a heterogeneous mention-level graph (hMG) to model complex interaction among different mentions across the document. It also constructs an entity-level graph (EG), based on which we propose a novel path reasoning mechanism to infer relations between entities. Experiments on the public dataset, DocRED, show GAIN achieves a significant performance improvement (2.85 on F1) over the previous state-of-the-art. Our code is available at https://github.com/DreamInvoker/GAIN .

cs.CL↗

arXiv2020

Cross-Modal Alignment with Mixture Experts Neural Network for Intral-City Retail Recommendation

In this paper, we introduce Cross-modal Alignment with mixture experts Neural Network (CameNN) recommendation model for intral-city retail industry, which aims to provide fresh foods and groceries retailing within 5 hours delivery service arising for the outbreak of Coronavirus disease (COVID-19) pandemic around the world. We propose CameNN, which is a multi-task model with three tasks including Image to Text Alignment (ITA) task, Text to Image Alignment (TIA) task and CVR prediction task. We use pre-trained BERT to generate the text embedding and pre-trained InceptionV4 to generate image patch embedding (each image is split into small patches with the same pixels and treat each patch as an image token). Softmax gating networks follow to learn the weight of each transformer expert output and choose only a subset of experts conditioned on the input. Then transformer encoder is applied as the share-bottom layer to learn all input features' shared interaction. Next, mixture of transformer experts (MoE) layer is implemented to model different aspects of tasks. At top of the MoE layer, we deploy a transformer layer for each task as task tower to learn task-specific information. On the real word intra-city dataset, experiments demonstrate CameNN outperform baselines and achieve significant improvements on the image and text representation. In practice, we applied CameNN on CVR prediction in our intra-city recommender system which is one of the leading intra-city platforms operated in China.

cs.CV↗

arXiv2020

Random Style Transfer based Domain Generalization Networks Integrating Shape and Spatial Information

Deep learning (DL)-based models have demonstrated good performance in medical image segmentation. However, the models trained on a known dataset often fail when performed on an unseen dataset collected from different centers, vendors and disease populations. In this work, we present a random style transfer network to tackle the domain generalization problem for multi-vendor and center cardiac image segmentation. Style transfer is used to generate training data with a wider distribution/ heterogeneity, namely domain augmentation. As the target domain could be unknown, we randomly generate a modality vector for the target modality in the style transfer stage, to simulate the domain shift for unknown domains. The model can be trained in a semi-supervised manner by simultaneously optimizing a supervised segmentation and an unsupervised style translation objective. Besides, the framework incorporates the spatial information and shape prior of the target by introducing two regularization terms. We evaluated the proposed framework on 40 subjects from the M\&Ms challenge2020, and obtained promising performance in the segmentation for data from unknown vendors and centers.

cs.CV↗

arXiv2020

Dispersed Exponential Family Mixture VAEs for Interpretable Text Generation

Deep generative models are commonly used for generating images and text. Interpretability of these models is one important pursuit, other than the generation quality. Variational auto-encoder (VAE) with Gaussian distribution as prior has been successfully applied in text generation, but it is hard to interpret the meaning of the latent variable. To enhance the controllability and interpretability, one can replace the Gaussian prior with a mixture of Gaussian distributions (GM-VAE), whose mixture components could be related to hidden semantic aspects of data. In this paper, we generalize the practice and introduce DEM-VAE, a class of models for text generation using VAEs with a mixture distribution of exponential family. Unfortunately, a standard variational training algorithm fails due to the mode-collapse problem. We theoretically identify the root cause of the problem and propose an effective algorithm to train DEM-VAE. Our method penalizes the training with an extra dispersion term to induce a well-structured latent space. Experimental results show that our approach does obtain a meaningful space, and it outperforms strong baselines in text generation benchmarks. The code is available at https://github.com/wenxianxian/demvae.

cs.LG↗

arXiv2020

Cross-Modality Multi-Atlas Segmentation Using Deep Neural Networks

Both image registration and label fusion in the multi-atlas segmentation (MAS) rely on the intensity similarity between target and atlas images. However, such similarity can be problematic when target and atlas images are acquired using different imaging protocols. High-level structure information can provide reliable similarity measurement for cross-modality images when cooperating with deep neural networks (DNNs). This work presents a new MAS framework for cross-modality images, where both image registration and label fusion are achieved by DNNs. For image registration, we propose a consistent registration network, which can jointly estimate forward and backward dense displacement fields (DDFs). Additionally, an invertible constraint is employed in the network to reduce the correspondence ambiguity of the estimated DDFs. For label fusion, we adapt a few-shot learning network to measure the similarity of atlas and target patches. Moreover, the network can be seamlessly integrated into the patch-based label fusion. The proposed framework is evaluated on the MM-WHS dataset of MICCAI 2017. Results show that the framework is effective in both cross-modality registration and segmentation.

cs.CV↗

arXiv2020

SketchDesc: Learning Local Sketch Descriptors for Multi-view Correspondence

In this paper, we study the problem of multi-view sketch correspondence, where we take as input multiple freehand sketches with different views of the same object and predict as output the semantic correspondence among the sketches. This problem is challenging since the visual features of corresponding points at different views can be very different. To this end, we take a deep learning approach and learn a novel local sketch descriptor from data. We contribute a training dataset by generating the pixel-level correspondence for the multi-view line drawings synthesized from 3D shapes. To handle the sparsity and ambiguity of sketches, we design a novel multi-branch neural network that integrates a patch-based representation and a multi-scale strategy to learn the pixel-level correspondence among multi-view sketches. We demonstrate the effectiveness of our proposed approach with extensive experiments on hand-drawn sketches and multi-view line drawings rendered from multiple 3D shape datasets.

cs.CV↗

arXiv2020

Generative Adversarial Network-Based Sinogram Super-Resolution for Computed Tomography Imaging

Compared with the conventional 1*1 acquisition mode of projection in computed tomography (CT) image reconstruction, the 2*2 acquisition mode improves the collection efficiency of the projection and reduces the X-ray exposure time. However, the collected projection based on the 2*2 acquisition mode has low resolution (LR) and the reconstructed image quality is poor, thus limiting the use of this mode in CT imaging systems. In this study, a novel sinogram-super-resolution generative adversarial network (SSR-GAN) model is proposed to obtain high-resolution (HR) sinograms from LR sinograms, thereby improving the reconstruction image quality under the 2*2 acquisition mode. The proposed generator is based on the residual network for LR sinogram feature extraction and super-resolution (SR) sinogram generation. A relativistic discriminator is designed to render the network capable of obtaining more realistic SR sinograms. Moreover, we combine the cycle consistency loss, sinogram domain loss, and reconstruction image domain loss in the total loss function to supervise SR sinogram generation. Then, a trained model can be obtained by inputting the paired LR/HR sinograms into the network. Finally, the classic FBP reconstruction algorithm is used for CT image reconstruction based on the generated SR sinogram. The qualitative and quantitative results of evaluations on digital and real data illustrate that the proposed model not only obtains clean SR sinograms from noisy LR sinograms but also outperforms its counterparts.

physics.med-ph↗