arXiv subjects

Lijun Wu

Publications and source records attributed to Lijun Wu.

At least 127 records · Page 7Linked to original sources

arXiv2021

R-Drop: Regularized Dropout for Neural Networks

Dropout is a powerful and widely used technique to regularize the training of deep neural networks. In this paper, we introduce a simple regularization strategy upon dropout in model training, namely R-Drop, which forces the output distributions of different sub models generated by dropout to be consistent with each other. Specifically, for each training sample, R-Drop minimizes the bidirectional KL-divergence between the output distributions of two sub models sampled by dropout. Theoretical analysis reveals that R-Drop reduces the freedom of the model parameters and complements dropout. Experiments on $\bf{5}$ widely used deep learning tasks ($\bf{18}$ datasets in total), including neural machine translation, abstractive summarization, language understanding, language modeling, and image classification, show that R-Drop is universally effective. In particular, it yields substantial improvements when applied to fine-tune large-scale pre-trained models, e.g., ViT, RoBERTa-large, and BART, and achieves state-of-the-art (SOTA) performances with the vanilla Transformer model on WMT14 English$\to$German translation ($\bf{30.91}$ BLEU) and WMT14 English$\to$French translation ($\bf{43.95}$ BLEU), even surpassing models trained with extra large-scale data and expert-designed advanced variants of Transformer models. Our code is available at GitHub{\url{https://github.com/dropreg/R-Drop}}.

cs.LG↗

arXiv2021

Pre-training Co-evolutionary Protein Representation via A Pairwise Masked Language Model

Understanding protein sequences is vital and urgent for biology, healthcare, and medicine. Labeling approaches are expensive yet time-consuming, while the amount of unlabeled data is increasing quite faster than that of the labeled data due to low-cost, high-throughput sequencing methods. In order to extract knowledge from these unlabeled data, representation learning is of significant value for protein-related tasks and has great potential for helping us learn more about protein functions and structures. The key problem in the protein sequence representation learning is to capture the co-evolutionary information reflected by the inter-residue co-variation in the sequences. Instead of leveraging multiple sequence alignment as is usually done, we propose a novel method to capture this information directly by pre-training via a dedicated language model, i.e., Pairwise Masked Language Model (PMLM). In a conventional masked language model, the masked tokens are modeled by conditioning on the unmasked tokens only, but processed independently to each other. However, our proposed PMLM takes the dependency among masked tokens into consideration, i.e., the probability of a token pair is not equal to the product of the probability of the two tokens. By applying this model, the pre-trained encoder is able to generate a better representation for protein sequences. Our result shows that the proposed method can effectively capture the inter-residue correlations and improves the performance of contact prediction by up to 9% compared to the MLM baseline under the same setting. The proposed model also significantly outperforms the MSA baseline by more than 7% on the TAPE contact prediction benchmark when pre-trained on a subset of the sequence database which the MSA is generated from, revealing the potential of the sequence pre-training method to surpass MSA based methods in general.

cs.CL↗

arXiv2021

How to Leverage Multimodal EHR Data for Better Medical Predictions?

Healthcare is becoming a more and more important research topic recently. With the growing data in the healthcare domain, it offers a great opportunity for deep learning to improve the quality of medical service. However, the complexity of electronic health records (EHR) data is a challenge for the application of deep learning. Specifically, the data produced in the hospital admissions are monitored by the EHR system, which includes structured data like daily body temperature, and unstructured data like free text and laboratory measurements. Although there are some preprocessing frameworks proposed for specific EHR data, the clinical notes that contain significant clinical value are beyond the realm of their consideration. Besides, whether these different data from various views are all beneficial to the medical tasks and how to best utilize these data remain unclear. Therefore, in this paper, we first extract the accompanying clinical notes from EHR and propose a method to integrate these data, we also comprehensively study the different models and the data leverage methods for better medical task prediction. The results on two medical prediction tasks show that our fused model with different data outperforms the state-of-the-art method that without clinical notes, which illustrates the importance of our fusion method and the value of clinical note features. Our code is available at https: //github.com/emnlp-mimic/mimic.

cs.CL↗

arXiv2021

Photoinduced anisotropic lattice dynamic response and domain formation in thermoelectric SnSe

Identifying and understanding the mechanisms behind strong phonon-phonon scattering in condensed matter systems is critical to maximizing the efficiency of thermoelectric devices. To date, the leading method to address this has been to meticulously survey the full phonon dispersion of the material in order to isolate modes with anomalously large linewidth and temperature-dependence. Here we combine quantitative MeV ultrafast electron diffraction (UED) analysis with Monte Carlo based dynamic diffraction simulation and first-principles calculations to directly unveil the soft, anharmonic lattice distortions of model thermoelectric material SnSe. A small single-crystal sample is photoexcited with ultrafast optical pulses and the soft, anharmonic lattice distortions are isolated using MeV-UED as those associated with long relaxation time and large displacements. We reveal that these modes have interlayer shear strain character, induced mainly by c-axis atomic displacements, resulting in domain formation in the transient state. These findings provide an innovative approach to identify mechanisms for ultralow and anisotropic thermal conductivity and a promising route to optimizing thermoelectric devices.

cond-mat.mtrl-sci↗

arXiv2021

Discovering Drug-Target Interaction Knowledge from Biomedical Literature

The Interaction between Drugs and Targets (DTI) in human body plays a crucial role in biomedical science and applications. As millions of papers come out every year in the biomedical domain, automatically discovering DTI knowledge from biomedical literature, which are usually triplets about drugs, targets and their interaction, becomes an urgent demand in the industry. Existing methods of discovering biological knowledge are mainly extractive approaches that often require detailed annotations (e.g., all mentions of biological entities, relations between every two entity mentions, etc.). However, it is difficult and costly to obtain sufficient annotations due to the requirement of expert knowledge from biomedical domains. To overcome these difficulties, we explore the first end-to-end solution for this task by using generative approaches. We regard the DTI triplets as a sequence and use a Transformer-based model to directly generate them without using the detailed annotations of entities and relations. Further, we propose a semi-supervised method, which leverages the aforementioned end-to-end model to filter unlabeled literature and label them. Experimental results show that our method significantly outperforms extractive baselines on DTI discovery. We also create a dataset, KD-DTI, to advance this task and will release it to the community.

cs.CL↗

arXiv2021

Signature of Many-Body Localization of Phonons in Strongly Disordered Superlattices

Many-body localization (MBL) has attracted significant attention due to its immunity to thermalization, role in logarithmic entanglement entropy growth, and opportunities to reach exotic quantum orders. However, experimental realization of MBL in solid-state systems has remained challenging. Here we report evidence of a possible phonon MBL phase in disordered GaAs/AlAs superlattices. Through grazing-incidence inelastic X-ray scattering, we observe a strong deviation of the phonon population from equilibrium in samples doped with ErAs nanodots at low temperature, signaling a departure from thermalization. This behavior occurs within finite phonon energy and wavevector windows, suggesting a localization-thermalization crossover. We support our observation by proposing a theoretical model for the effective phonon Hamiltonian in disordered superlattices, and showing that it can be mapped exactly to a disordered 1D Bose-Hubbard model with a known MBL phase. Our work provides momentum-resolved experimental evidence of phonon localization, extending the scope of MBL to disordered solid-state systems.

cond-mat.mtrl-sci↗

arXiv2021

Temporally Correlated Task Scheduling for Sequence Learning

Sequence learning has attracted much research attention from the machine learning community in recent years. In many applications, a sequence learning task is usually associated with multiple temporally correlated auxiliary tasks, which are different in terms of how much input information to use or which future step to predict. For example, (i) in simultaneous machine translation, one can conduct translation under different latency (i.e., how many input words to read/wait before translation); (ii) in stock trend forecasting, one can predict the price of a stock in different future days (e.g., tomorrow, the day after tomorrow). While it is clear that those temporally correlated tasks can help each other, there is a very limited exploration on how to better leverage multiple auxiliary tasks to boost the performance of the main task. In this work, we introduce a learnable scheduler to sequence learning, which can adaptively select auxiliary tasks for training depending on the model status and the current training data. The scheduler and the model for the main task are jointly trained through bi-level optimization. Experiments show that our method significantly improves the performance of simultaneous machine translation and stock trend forecasting.

cs.CL↗

arXiv2021

UniDrop: A Simple yet Effective Technique to Improve Transformer without Extra Cost

Transformer architecture achieves great success in abundant natural language processing tasks. The over-parameterization of the Transformer model has motivated plenty of works to alleviate its overfitting for superior performances. With some explorations, we find simple techniques such as dropout, can greatly boost model performance with a careful design. Therefore, in this paper, we integrate different dropout techniques into the training of Transformer models. Specifically, we propose an approach named UniDrop to unites three different dropout techniques from fine-grain to coarse-grain, i.e., feature dropout, structure dropout, and data dropout. Theoretically, we demonstrate that these three dropouts play different roles from regularization perspectives. Empirically, we conduct experiments on both neural machine translation and text classification benchmark datasets. Extensive results indicate that Transformer with UniDrop can achieve around 1.5 BLEU improvement on IWSLT14 translation tasks, and better accuracy for the classification even using strong pre-trained RoBERTa as backbone.

cs.CL↗

arXiv2021

IOT: Instance-wise Layer Reordering for Transformer Structures

With sequentially stacked self-attention, (optional) encoder-decoder attention, and feed-forward layers, Transformer achieves big success in natural language processing (NLP), and many variants have been proposed. Currently, almost all these models assume that the layer order is fixed and kept the same across data samples. We observe that different data samples actually favor different orders of the layers. Based on this observation, in this work, we break the assumption of the fixed layer order in the Transformer and introduce instance-wise layer reordering into the model structure. Our Instance-wise Ordered Transformer (IOT) can model variant functions by reordered layers, which enables each sample to select the better one to improve the model performance under the constraint of almost the same number of parameters. To achieve this, we introduce a light predictor with negligible parameter and inference cost to decide the most capable and favorable layer order for any input sequence. Experiments on 3 tasks (neural machine translation, abstractive summarization, and code generation) and 9 datasets demonstrate consistent improvements of our method. We further show that our method can also be applied to other architectures beyond Transformer. Our code is released at Github.

cs.CL↗

arXiv2021

Learning to Reweight with Deep Interactions

Recently, the concept of teaching has been introduced into machine learning, in which a teacher model is used to guide the training of a student model (which will be used in real tasks) through data selection, loss function design, etc. Learning to reweight, which is a specific kind of teaching that reweights training data using a teacher model, receives much attention due to its simplicity and effectiveness. In existing learning to reweight works, the teacher model only utilizes shallow/surface information such as training iteration number and loss/accuracy of the student model from training/validation sets, but ignores the internal states of the student model, which limits the potential of learning to reweight. In this work, we propose an improved data reweighting algorithm, in which the student model provides its internal states to the teacher model, and the teacher model returns adaptive weights of training samples to enhance the training of the student model. The teacher model is jointly trained with the student model using meta gradients propagated from a validation set. Experiments on image classification with clean/noisy labels and neural machine translation empirically demonstrate that our algorithm makes significant improvement over previous methods.

cs.LG↗

arXiv2020

ADD: Augmented Disentanglement Distillation Framework for Improving Stock Trend Forecasting

Stock trend forecasting has become a popular research direction that attracts widespread attention in the financial field. Though deep learning methods have achieved promising results, there are still many limitations, for example, how to extract clean features from the raw stock data. In this paper, we introduce an \emph{Augmented Disentanglement Distillation (ADD)} approach to remove interferential features from the noised raw data. Specifically, we present 1) a disentanglement structure to separate excess and market information from the stock data to avoid the two factors disturbing each other's own prediction. Besides, by applying 2) a dynamic self-distillation method over the disentanglement framework, other implicit interference factors can also be removed. Further, thanks to the decoder module in our framework, 3) a novel strategy is proposed to augment the training samples based on the different excess and market features to improve performance. We conduct experiments on the Chinese stock market data. Results show that our method significantly improves the stock trend forecasting performances, as well as the actual investment income through backtesting, which strongly demonstrates the effectiveness of our approach.

cs.LG↗

arXiv2020

Masked Contrastive Representation Learning for Reinforcement Learning

Improving sample efficiency is a key research problem in reinforcement learning (RL), and CURL, which uses contrastive learning to extract high-level features from raw pixels of individual video frames, is an efficient algorithm~\citep{srinivas2020curl}. We observe that consecutive video frames in a game are highly correlated but CURL deals with them independently. To further improve data efficiency, we propose a new algorithm, masked contrastive representation learning for RL, that takes the correlation among consecutive inputs into consideration. In addition to the CNN encoder and the policy network in CURL, our method introduces an auxiliary Transformer module to leverage the correlations among video frames. During training, we randomly mask the features of several frames, and use the CNN encoder and Transformer to reconstruct them based on the context frames. The CNN encoder and Transformer are jointly trained via contrastive learning where the reconstructed features should be similar to the ground-truth ones while dissimilar to others. During inference, the CNN encoder and the policy network are used to take actions, and the Transformer module is discarded. Our method achieves consistent improvements over CURL on $14$ out of $16$ environments from DMControl suite and $21$ out of $26$ environments from Atari 2600 Games. The code is available at https://github.com/teslacool/m-curl.

cs.LG↗

arXiv2020

Vacancy defect control of colossal thermopower in FeSb2

Iron diantimonide is a material with the highest known thermoelectric power. By combining scanning transmission electron microscope study with electronic transport neutron, X-ray scattering and first principle calculation we identify atomic defects that control colossal thermopower magnitude and nanoprecipitate clusters with Sb vacancy ordering which induce additional phonon scattering and substantially reduce thermal conductivity. Defects are found to cause rather weak but important monoclinic distortion of the unit cell Pnnm to Pm. The absence of Sb along [010] for high defect concentration forms conducting path due to Fe d orbital overlap. The connection between atomic defect anisotropy and colossal thermopower in FeSb2 paves the way for the understanding and tailoring of giant thermopower in related materials.

cond-mat.mtrl-sci↗

arXiv2020

Multi-branch Attentive Transformer

While the multi-branch architecture is one of the key ingredients to the success of computer vision tasks, it has not been well investigated in natural language processing, especially sequence learning tasks. In this work, we propose a simple yet effective variant of Transformer called multi-branch attentive Transformer (briefly, MAT), where the attention layer is the average of multiple branches and each branch is an independent multi-head attention layer. We leverage two training techniques to regularize the training: drop-branch, which randomly drops individual branches during training, and proximal initialization, which uses a pre-trained Transformer model to initialize multiple branches. Experiments on machine translation, code generation and natural language understanding demonstrate that such a simple variant of Transformer brings significant improvements. Our code is available at \url{https://github.com/HA-Transformer}.

cs.CL↗

arXiv2020

Water-induced formation of an alkali-ion dimer in cryptomelane nanorods

Tunneled metal oxides such as $α-$Mn$_8$O$_{16}$ (hollandite) have proven to be compelling candidates for charge-storage materials in high-density batteries. In particular, the tunnels can support one-dimensional chains of K$^+$ ions (which act as structure-stabilizing dopants) and H$_2$O molecules, as these chains are favored by strong H-bonds and electrostatic interactions. In this work, we examine the role of water molecules in enhancing the stability of K$^+$-doped $α-$Mn$_8$O$_{16}$ (cryptomelane). The combined experimental and theoretical analyses show that for high enough concentrations of water and tunnel-ions, H$_2$O displaces K$^+$ ions from their natural binding sites. This displacement becomes energetically favorable due to the formation of K$_2^+$ dimers, thereby modifying the stoichiometric charge of the system. These findings have potentially significant technological implications for the consideration of cryptomelane as a Li$^+$/Na$^+$ battery electrode. Our work establishes the functional role of water in altering the energetics and structural properties of cryptomelane, an observation that has frequently been overlooked in previous studies.

cond-mat.mtrl-sci↗

arXiv2020

Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection

One-stage detector basically formulates object detection as dense classification and localization. The classification is usually optimized by Focal Loss and the box location is commonly learned under Dirac delta distribution. A recent trend for one-stage detectors is to introduce an individual prediction branch to estimate the quality of localization, where the predicted quality facilitates the classification to improve detection performance. This paper delves into the representations of the above three fundamental elements: quality estimation, classification and localization. Two problems are discovered in existing practices, including (1) the inconsistent usage of the quality estimation and classification between training and inference and (2) the inflexible Dirac delta distribution for localization when there is ambiguity and uncertainty in complex scenes. To address the problems, we design new representations for these elements. Specifically, we merge the quality estimation into the class prediction vector to form a joint representation of localization quality and classification, and use a vector to represent arbitrary distribution of box locations. The improved representations eliminate the inconsistency risk and accurately depict the flexible distribution in real data, but contain continuous labels, which is beyond the scope of Focal Loss. We then propose Generalized Focal Loss (GFL) that generalizes Focal Loss from its discrete form to the continuous version for successful optimization. On COCO test-dev, GFL achieves 45.0\% AP using ResNet-101 backbone, surpassing state-of-the-art SAPD (43.5\%) and ATSS (43.6\%) with higher or comparable inference speed, under the same backbone and training settings. Notably, our best model can achieve a single-model single-scale AP of 48.2\%, at 10 FPS on a single 2080Ti GPU. Code and models are available at https://github.com/implus/GFocal.

cs.CV↗

arXiv2020

Concurrent probing of electron-lattice dephasing induced by photoexcitation in 1T-TaSeTe using ultrafast electron diffraction

It has been technically challenging to concurrently probe the electrons and the lattices in materials during non-equilibrium processes, allowing their correlations to be determined. Here, in a single set of ultrafast electron diffraction patterns taken on the charge-density-wave (CDW) material 1T-TaSeTe, we discover a temporal shift in the diffraction intensity measurements as a function of scattering angle. With the help of dynamic models and theoretical calculations, we show that the ultrafast electrons probe both the valence-electron and lattice dynamic processes, resulting in the temporal shift measurements. Our results demonstrate unambiguously that the CDW is not merely a result of the periodic lattice deformation ever-present in 1T-TaSeTe but has significant electronic origin. This method demonstrates a novel approach for studying many quantum effects that arise from electron-lattice dephasing in molecules and crystals for next-generation devices.

cond-mat.mtrl-sci↗

arXiv2020

Incorporating BERT into Neural Machine Translation

The recently proposed BERT has shown great power on a variety of natural language understanding tasks, such as text classification, reading comprehension, etc. However, how to effectively apply BERT to neural machine translation (NMT) lacks enough exploration. While BERT is more commonly used as fine-tuning instead of contextual embedding for downstream language understanding tasks, in NMT, our preliminary exploration of using BERT as contextual embedding is better than using for fine-tuning. This motivates us to think how to better leverage BERT for NMT along this direction. We propose a new algorithm named BERT-fused model, in which we first use BERT to extract representations for an input sequence, and then the representations are fused with each layer of the encoder and decoder of the NMT model through attention mechanisms. We conduct experiments on supervised (including sentence-level and document-level translations), semi-supervised and unsupervised machine translation, and achieve state-of-the-art results on seven benchmark datasets. Our code is available at \url{https://github.com/bert-nmt/bert-nmt}.

cs.CL↗