arXiv subjects

Jin Xu

Publications and source records attributed to Jin Xu.

At least 145 records · Page 8Linked to original sources

arXiv2023

Deep Stochastic Processes via Functional Markov Transition Operators

We introduce Markov Neural Processes (MNPs), a new class of Stochastic Processes (SPs) which are constructed by stacking sequences of neural parameterised Markov transition operators in function space. We prove that these Markov transition operators can preserve the exchangeability and consistency of SPs. Therefore, the proposed iterative construction adds substantial flexibility and expressivity to the original framework of Neural Processes (NPs) without compromising consistency or adding restrictions. Our experiments demonstrate clear advantages of MNPs over baseline models on a variety of tasks.

stat.ML↗

arXiv2023

DC-MBR: Distributional Cooling for Minimum Bayesian Risk Decoding

Minimum Bayesian Risk Decoding (MBR) emerges as a promising decoding algorithm in Neural Machine Translation. However, MBR performs poorly with label smoothing, which is surprising as label smoothing provides decent improvement with beam search and improves generality in various tasks. In this work, we show that the issue arises from the un-consistency of label smoothing on the token-level and sequence-level distributions. We demonstrate that even though label smoothing only causes a slight change in the token-level, the sequence-level distribution is highly skewed. We coin the issue \emph{autoregressive over-smoothness}. To address this issue, we propose a simple and effective method, Distributional Cooling MBR (DC-MBR), which manipulates the entropy of output distributions by tuning down the Softmax temperature. We theoretically prove the equivalence between pre-tuning label smoothing factor and distributional cooling. Extensive experiments on NMT benchmarks validate that distributional cooling improves MBR in various settings.

cs.CL↗

arXiv2023

How should the server sleep? -- Age-energy tradeoff in sleep-wake server systems

The surging demand for fresh information from various Internet of Things (IoT) applications requires oceans of data sampled from the physical environment to be transmitted and processed timely, which would lead to huge energy consumption. We investigate the sleep-wake strategies of servers in communication systems to reduce energy consumption while guaranteeing timely delivery of fresh information to users. Specifically, we investigate a multi-source single-server queueing system and propose a novel sleep-wake strategy called the Conditional Sleep (CS) scheme. Our analysis reveals that the CS scheme outperforms the widely-used Hysteresis Time (HT) and Bernoulli Sleep (BS) schemes in terms of Age of Information (AoI), while retaining the same energy consumption rate and Peak Age of Information (PAoI). We find that increasing the sleep period length leads to a reduction in energy consumption and an increase in PAoI, but it does not always increase AoI. Moreover, we show that using PAoI as the information freshness metric in designing sleep-wake strategies would make the server sleep infinitely long due to the PAoI being determined by first-order statistics. We further numerically show that having the bufferless system can achieve a better PAoI-energy tradeoff than the infinite buffer system when having a large sampling rate.

cs.IT↗

arXiv2023

Investigating Graph Structure Information for Entity Alignment with Dangling Cases

Entity alignment (EA) aims to discover the equivalent entities in different knowledge graphs (KGs), which play an important role in knowledge engineering. Recently, EA with dangling entities has been proposed as a more realistic setting, which assumes that not all entities have corresponding equivalent entities. In this paper, we focus on this setting. Some work has explored this problem by leveraging translation API, pre-trained word embeddings, and other off-the-shelf tools. However, these approaches over-rely on the side information (e.g., entity names), and fail to work when the side information is absent. On the contrary, they still insufficiently exploit the most fundamental graph structure information in KG. To improve the exploitation of the structural information, we propose a novel entity alignment framework called Weakly-Optimal Graph Contrastive Learning (WOGCL), which is refined on three dimensions : (i) Model. We propose a novel Gated Graph Attention Network to capture local and global graph structure similarity. (ii) Training. Two learning objectives: contrastive learning and optimal transport learning are designed to obtain distinguishable entity representations via the optimal transport plan. (iii) Inference. In the inference phase, a PageRank-based method is proposed to calculate higher-order structural similarity. Extensive experiments on two dangling benchmarks demonstrate that our WOGCL outperforms the current state-of-the-art methods with pure structural information in both traditional (relaxed) and dangling (consolidated) settings. The code will be public soon.

cs.CL↗

arXiv2022

Simultaneous control of spectral and directional emissivity with gradient epsilon-near-zero InAs photonic structures

Controlling both the spectral bandwidth and directional range of emitted thermal radiation is a fundamental challenge in modern photonics and materials research. Recent work has shown that materials with a spatial gradient in their epsilon near zero response can support broad spectrum directionality in their emissivity, enabling high radiance to specific angles of incidence. However, this capability has been limited spectrally and directionally by the availability of materials supporting phonon-polariton resonances over long-wave infrared wavelengths. Here, we design and experimentally demonstrate an approach using doped III-V semiconductors that can simultaneously tailor spectral peak, bandwidth and directionality of infrared emissivity. We epitaxially grow and characterize InAs-based gradient ENZ photonic structures that exhibit broadband directional emission with varying spectral bandwidths and peak directions as a function of their doping concentration profile and thickness. Due to its easy-to-fabricate geometry we believe this approach provides a versatile photonic platform to dynamically control broadband spectral and directional emissivity for a range of emerging applications.

physics.optics↗

arXiv2022

FastCorrect: Fast Error Correction with Edit Alignment for Automatic Speech Recognition

Error correction techniques have been used to refine the output sentences from automatic speech recognition (ASR) models and achieve a lower word error rate (WER) than original ASR outputs. Previous works usually use a sequence-to-sequence model to correct an ASR output sentence autoregressively, which causes large latency and cannot be deployed in online ASR services. A straightforward solution to reduce latency, inspired by non-autoregressive (NAR) neural machine translation, is to use an NAR sequence generation model for ASR error correction, which, however, comes at the cost of significantly increased ASR error rate. In this paper, observing distinctive error patterns and correction operations (i.e., insertion, deletion, and substitution) in ASR, we propose FastCorrect, a novel NAR error correction model based on edit alignment. In training, FastCorrect aligns each source token from an ASR output sentence to the target tokens from the corresponding ground-truth sentence based on the edit distance between the source and target sentences, and extracts the number of target tokens corresponding to each source token during edition/correction, which is then used to train a length predictor and to adjust the source tokens to match the length of the target sentence for parallel generation. In inference, the token number predicted by the length predictor is used to adjust the source tokens for target sequence generation. Experiments on the public AISHELL-1 dataset and an internal industrial-scale ASR dataset show the effectiveness of FastCorrect for ASR error correction: 1) it speeds up the inference by 6-9 times and maintains the accuracy (8-14% WER reduction) compared with the autoregressive correction model; and 2) it outperforms the popular NAR models adopted in neural machine translation and text edition by a large margin.

cs.CL↗

arXiv2022

FastCorrect 2: Fast Error Correction on Multiple Candidates for Automatic Speech Recognition

Error correction is widely used in automatic speech recognition (ASR) to post-process the generated sentence, and can further reduce the word error rate (WER). Although multiple candidates are generated by an ASR system through beam search, current error correction approaches can only correct one sentence at a time, failing to leverage the voting effect from multiple candidates to better detect and correct error tokens. In this work, we propose FastCorrect 2, an error correction model that takes multiple ASR candidates as input for better correction accuracy. FastCorrect 2 adopts non-autoregressive generation for fast inference, which consists of an encoder that processes multiple source sentences and a decoder that generates the target sentence in parallel from the adjusted source sentence, where the adjustment is based on the predicted duration of each source token. However, there are some issues when handling multiple source sentences. First, it is non-trivial to leverage the voting effect from multiple source sentences since they usually vary in length. Thus, we propose a novel alignment algorithm to maximize the degree of token alignment among multiple sentences in terms of token and pronunciation similarity. Second, the decoder can only take one adjusted source sentence as input, while there are multiple source sentences. Thus, we develop a candidate predictor to detect the most suitable candidate for the decoder. Experiments on our inhouse dataset and AISHELL-1 show that FastCorrect 2 can further reduce the WER over the previous correction model with single candidate by 3.2% and 2.6%, demonstrating the effectiveness of leveraging multiple candidates in ASR error correction. FastCorrect 2 achieves better performance than the cascaded re-scoring and correction pipeline and can serve as a unified post-processing module for ASR.

cs.CL↗

arXiv2022

A universal DNA computing model for solving NP-hard subset problems

DNA computing, a nontraditional computing mechanism, provides a feasible and effective method for solving NP-hard problems because of the vast parallelism and high-density storage of DNA molecules. Although DNA computing has been exploited to solve various intractable computational problems, such as the Hamiltonian path problem, SAT problem, and graph coloring problem, there has been little discussion of designing universal DNA computing-based models, which can solve a class of problems. In this paper, by leveraging the dynamic and enzyme-free properties of DNA strand displacement, we propose a universal model named DCMSubset for solving subset problems in graph theory. The model aims to find a minimum (or maximum) set satisfying given constraints. For each element x involved in a given problem, DCMSubset uses an exclusive single-stranded DNA molecule to model x as well as a specific DNA complex to model the relationship between x and other elements. Based on the proposed model, we conducted simulation and biochemical experiments on three kinds of subset problems, a minimum dominating set, maximum independent set, and minimum vertex cover. We observed that DCMSubset can also be used to solve the graph coloring problem. Moreover, we extended DCMSubset to a model for solving the SAT problem. The results of experiments showed the feasibility and university of the proposed method. Our results highlighted the potential for DNA strand displacement to act as a computation tool to solve NP-hard problems.

q-bio.MN↗

arXiv2022

SBMLDiagrams: A python package to process and visualize SBML layout and render

Summary: The Systems Biology Markup Language (SBML) is an extensible standard format for exchanging biochemical models. One of the extensions for SBML is the SBML Layout and Render package. This allows modelers to describe a biochemical model as a pathway diagram. However, up to now there has been little support to help users easily add and retrieve such information from SBML. In this application note, we describe a new Python package called SBMLDiagrams. This package allows a user to add layout and render information or retrieve it using a straightforward Python API. The package uses skia-python to support the rendering of the diagrams, allowing export to commons formats such as PNG or PDF. Availability: SBMLDiagrams is publicly available and licensed under the liberal MIT open-source license. The package is available for all major platforms. The source code has been deposited at GitHub (github.com/sys-bio/SBMLDiagrams). Users can install the package using the standard pip installation mechanism: pip install SBMLDiagrams. Contact: hsauro@uw.edu.

q-bio.QM↗

arXiv2022

Overlapping time of a virtual customer in time-varying many-server queues

Motivated by the ongoing COVID-19 pandemic, this paper investigates customers' infection risk by evaluating the overlapping time of a virtual customer with others in queueing systems. Most of the current methodologies focus on characterizing the risk in stationary systems, which may not apply to the more practical time-varying systems. As such, we propose an approximation framework that relies on the fluid limit to compute the expected overlapping time in time-varying queueing systems. Simulation experiments verify the accuracy of our approach.

math.PR↗

arXiv2022

$N$-gram Is Back: Residual Learning of Neural Text Generation with $n$-gram Language Model

$N$-gram language models (LM) have been largely superseded by neural LMs as the latter exhibits better performance. However, we find that $n$-gram models can achieve satisfactory performance on a large proportion of testing cases, indicating they have already captured abundant knowledge of the language with relatively low computational cost. With this observation, we propose to learn a neural LM that fits the residual between an $n$-gram LM and the real-data distribution. The combination of $n$-gram and neural LMs not only allows the neural part to focus on the deeper understanding of language but also provides a flexible way to customize an LM by switching the underlying $n$-gram model without changing the neural model. Experimental results on three typical language tasks (i.e., language modeling, machine translation, and summarization) demonstrate that our approach attains additional performance gains over popular standalone neural models consistently. We also show that our approach allows for effective domain adaptation by simply switching to a domain-specific $n$-gram model, without any extra training. Our code is released at https://github.com/ghrua/NgramRes.

cs.CL↗

arXiv2022

A comparative study to alternatives to the log-rank test

Studies to compare the survival of two or more groups using time-to-event data are of high importance in medical research. The gold standard is the log-rank test, which is optimal under proportional hazards. As the latter is no simple regularity assumption, we are interested in evaluating the power of various statistical tests under different settings including proportional and non-proportional hazards with a special emphasize on crossing hazards. This challenge has been going on for many years now and multiple methods have already been investigated in extensive simulation studies. However, in recent years new omnibus tests and methods based on the restricted mean survival time appeared that have been strongly recommended in biometric literature. Thus, to give updated recommendations, we perform a vast simulation study to compare tests that showed high power in previous studies with these more recent approaches. We thereby analyze various simulation settings with varying survival and censoring distributions, unequal censoring between groups, small sample sizes and unbalanced group sizes. Overall, omnibus tests are more robust in terms of power against deviations from the proportional hazards assumption.

stat.ME↗

arXiv2022

Learning to Break the Loop: Analyzing and Mitigating Repetitions for Neural Text Generation

While large-scale neural language models, such as GPT2 and BART, have achieved impressive results on various text generation tasks, they tend to get stuck in undesirable sentence-level loops with maximization-based decoding algorithms (\textit{e.g.}, greedy search). This phenomenon is counter-intuitive since there are few consecutive sentence-level repetitions in human corpora (e.g., 0.02\% in Wikitext-103). To investigate the underlying reasons for generating consecutive sentence-level repetitions, we study the relationship between the probabilities of the repetitive tokens and their previous repetitions in the context. Through our quantitative experiments, we find that 1) Language models have a preference to repeat the previous sentence; 2) The sentence-level repetitions have a \textit{self-reinforcement effect}: the more times a sentence is repeated in the context, the higher the probability of continuing to generate that sentence; 3) The sentences with higher initial probabilities usually have a stronger self-reinforcement effect. Motivated by our findings, we propose a simple and effective training method \textbf{DITTO} (Pseu\underline{D}o-Repet\underline{IT}ion Penaliza\underline{T}i\underline{O}n), where the model learns to penalize probabilities of sentence-level repetitions from pseudo repetitive data. Although our method is motivated by mitigating repetitions, experiments show that DITTO not only mitigates the repetition issue without sacrificing perplexity, but also achieves better generation quality. Extensive experiments on open-ended text generation (Wikitext-103) and text summarization (CNN/DailyMail) demonstrate the generality and effectiveness of our method.

cs.CL↗

arXiv2022

Analyzing and Mitigating Interference in Neural Architecture Search

Weight sharing is a popular approach to reduce the cost of neural architecture search (NAS) by reusing the weights of shared operators from previously trained child models. However, the rank correlation between the estimated accuracy and ground truth accuracy of those child models is low due to the interference among different child models caused by weight sharing. In this paper, we investigate the interference issue by sampling different child models and calculating the gradient similarity of shared operators, and observe: 1) the interference on a shared operator between two child models is positively correlated with the number of different operators; 2) the interference is smaller when the inputs and outputs of the shared operator are more similar. Inspired by these two observations, we propose two approaches to mitigate the interference: 1) MAGIC-T: rather than randomly sampling child models for optimization, we propose a gradual modification scheme by modifying one operator between adjacent optimization steps to minimize the interference on the shared operators; 2) MAGIC-A: forcing the inputs and outputs of the operator across all child models to be similar to reduce the interference. Experiments on a BERT search space verify that mitigating interference via each of our proposed methods improves the rank correlation of super-pet and combining both methods can achieve better results. Our discovered architecture outperforms RoBERTa$_{\rm base}$ by 1.1 and 0.6 points and ELECTRA$_{\rm base}$ by 1.6 and 1.1 points on the dev and test set of GLUE benchmark. Extensive results on the BERT compression, reading comprehension and ImageNet task demonstrate the effectiveness and generality of our proposed methods.

cs.CL↗

arXiv2022

AGIC: Approximate Gradient Inversion Attack on Federated Learning

Federated learning is a private-by-design distributed learning paradigm where clients train local models on their own data before a central server aggregates their local updates to compute a global model. Depending on the aggregation method used, the local updates are either the gradients or the weights of local learning models. Recent reconstruction attacks apply a gradient inversion optimization on the gradient update of a single minibatch to reconstruct the private data used by clients during training. As the state-of-the-art reconstruction attacks solely focus on single update, realistic adversarial scenarios are overlooked, such as observation across multiple updates and updates trained from multiple mini-batches. A few studies consider a more challenging adversarial scenario where only model updates based on multiple mini-batches are observable, and resort to computationally expensive simulation to untangle the underlying samples for each local step. In this paper, we propose AGIC, a novel Approximate Gradient Inversion Attack that efficiently and effectively reconstructs images from both model or gradient updates, and across multiple epochs. In a nutshell, AGIC (i) approximates gradient updates of used training samples from model updates to avoid costly simulation procedures, (ii) leverages gradient/model updates collected from multiple epochs, and (iii) assigns increasing weights to layers with respect to the neural network structure for reconstruction quality. We extensively evaluate AGIC on three datasets, CIFAR-10, CIFAR-100 and ImageNet. Our results show that AGIC increases the peak signal-to-noise ratio (PSNR) by up to 50% compared to two representative state-of-the-art gradient inversion attacks. Furthermore, AGIC is faster than the state-of-the-art simulation based attack, e.g., it is 5x faster when attacking FedAvg with 8 local steps in between model updates.

cs.LG↗

arXiv2022

A Novel Markov Model for Near-Term Railway Delay Prediction

Predicting the near-future delay with accuracy for trains is momentous for railway operations and passengers' traveling experience. This work aims to design prediction models for train delays based on Netherlands Railway data. We first develop a chi-square test to show that the delay evolution over stations follows a first-order Markov chain. We then propose a delay prediction model based on non-homogeneous Markov chains. To deal with the sparsity of the transition matrices of the Markov chains, we propose a novel matrix recovery approach that relies on Gaussian kernel density estimation. Our numerical tests show that this recovery approach outperforms other heuristic approaches in prediction accuracy. The Markov chain model we propose also shows to be better than other widely-used time series models with respect to both interpretability and prediction accuracy. Moreover, our proposed model does not require a complicated training process, which is capable of handling large-scale forecasting problems.

cs.LG↗

arXiv2022

Matrix Multiplication with Less Arithmetic Complexity and IO Complexity

After Strassen presented the first sub-cubic matrix multiplication algorithm, many Strassen-like algorithms are presented. Most of them with low asymptotic cost have large hidden leading coefficient which are thus impractical. To reduce the leading coefficient, Cenk and Hasan give a general approach reducing the leading coefficient of $<2,2,2;7>$-algorithm to $5$ but increasing IO complexity. In 2017, Karstadt and Schwartz also reduce the leading coefficient of $<2,2,2;7>$-algorithm to $5$ by the Alternative Basis Matrix Multiplication method. Meanwhile, their method reduces the IO complexity and low-order monomials in arithmetic complexity. In 2019, Beniamini and Schwartz generalize Alternative Basis Matrix Multiplication method reducing leading coefficient in arithmetic complexity but increasing IO complexity. In this paper, we propose a new matrix multiplication algorithm which reduces leading coefficient both in arithmetic complexity and IO complexity. We apply our method to Strassen-like algorithms improving arithmetic complexity and IO complexity (the comparison with previous results are shown in Tables 1 and 2). Surprisingly, our IO complexity of $<3,3,3;23>$-algorithm is $14n^{\log_323}M^{-\frac{1}{2}} + o(n^{\log_323})$ which breaks Ballard's IO complexity low bound ($Ω(n^{\log_323}M^{1-\frac{\log_323}{2}})$) for recursive Strassen-like algorithms.

cs.SC↗

arXiv2022

ECO v1: Towards Event-Centric Opinion Mining

Events are considered as the fundamental building blocks of the world. Mining event-centric opinions can benefit decision making, people communication, and social good. Unfortunately, there is little literature addressing event-centric opinion mining, although which significantly diverges from the well-studied entity-centric opinion mining in connotation, structure, and expression. In this paper, we propose and formulate the task of event-centric opinion mining based on event-argument structure and expression categorizing theory. We also benchmark this task by constructing a pioneer corpus and designing a two-step benchmark framework. Experiment results show that event-centric opinion mining is feasible and challenging, and the proposed task, dataset, and baselines are beneficial for future studies.

cs.CL↗