arXiv subjects

Yu Du

Publications and source records attributed to Yu Du.

At least 55 records · Page 3Linked to original sources

arXiv2023

Teeth And Root Canals Segmentation Using ZXYFormer With Uncertainty Guidance And Weight Transfer

This study attempts to segment teeth and root-canals simultaneously from CBCT images, but there are very challenging problems in this process. First, the clinical CBCT image data is very large (e.g., 672 *688 * 688), and the use of downsampling operation will lose useful information about teeth and root canals. Second, teeth and root canals are very different in morphology, and it is difficult for a simple network to identify them precisely. In addition, there are weak edges at the tooth, between tooth and root canal, which makes it very difficult to segment such weak edges. To this end, we propose a coarse-to-fine segmentation method based on inverse feature fusion transformer and uncertainty estimation to address above challenging problems. First, we use the downscaled volume data (e.g., 128 * 128 * 128) to conduct coarse segmentation and map it to the original volume to obtain the area of teeth and root canals. Then, we design a transformer with reverse feature fusion, which can bring better segmentation effect of different morphological objects by transferring deeper features to shallow features. Finally, we design an auxiliary branch to calculate and refine the difficult areas in order to improve the weak edge segmentation performance of teeth and root canals. Through the combined tooth and root canal segmentation experiment of 157 clinical high-resolution CBCT data, it is verified that the proposed method is superior to the existing tooth or root canal segmentation methods.

cs.CV↗

arXiv2023

Improving the mixed model for repeated measures to robustly increase precision in randomized trials

In randomized trials, repeated measures of the outcome are routinely collected. The mixed model for repeated measures (MMRM) leverages the information from these repeated outcome measures, and is often used for the primary analysis to estimate the average treatment effect at the primary endpoint. MMRM, however, can suffer from bias and precision loss when it models intermediate outcomes incorrectly, and hence fails to use the post-randomization information harmlessly. This paper proposes an extension of the commonly used MMRM, called IMMRM, that improves the robustness and optimizes the precision gain from covariate adjustment, stratified randomization, and adjustment for intermediate outcome measures. Under regularity conditions and missing completely at random, we prove that the IMMRM estimator for the average treatment effect is robust to arbitrary model misspecification and is asymptotically equal or more precise than the analysis of covariance (ANCOVA) estimator and the MMRM estimator. Under missing at random, IMMRM is less likely to be misspecified than MMRM, and we demonstrate via simulation studies that IMMRM continues to have less bias and smaller variance. Our results are further supported by a re-analysis of a randomized trial for the treatment of diabetes.

stat.ME↗

arXiv2023

Boosting Zero-shot Learning via Contrastive Optimization of Attribute Representations

Zero-shot learning (ZSL) aims to recognize classes that do not have samples in the training set. One representative solution is to directly learn an embedding function associating visual features with corresponding class semantics for recognizing new classes. Many methods extend upon this solution, and recent ones are especially keen on extracting rich features from images, e.g. attribute features. These attribute features are normally extracted within each individual image; however, the common traits for features across images yet belonging to the same attribute are not emphasized. In this paper, we propose a new framework to boost ZSL by explicitly learning attribute prototypes beyond images and contrastively optimizing them with attribute-level features within images. Besides the novel architecture, two elements are highlighted for attribute representations: a new prototype generation module is designed to generate attribute prototypes from attribute semantics; a hard example-based contrastive optimization scheme is introduced to reinforce attribute-level features in the embedding space. We explore two alternative backbones, CNN-based and transformer-based, to build our framework and conduct experiments on three standard benchmarks, CUB, SUN, AwA2. Results on these benchmarks demonstrate that our method improves the state of the art by a considerable margin. Our codes will be available at https://github.com/dyabel/CoAR-ZSL.git

cs.CV↗

arXiv2023

Coupling Artificial Neurons in BERT and Biological Neurons in the Human Brain

Linking computational natural language processing (NLP) models and neural responses to language in the human brain on the one hand facilitates the effort towards disentangling the neural representations underpinning language perception, on the other hand provides neurolinguistics evidence to evaluate and improve NLP models. Mappings of an NLP model's representations of and the brain activities evoked by linguistic input are typically deployed to reveal this symbiosis. However, two critical problems limit its advancement: 1) The model's representations (artificial neurons, ANs) rely on layer-level embeddings and thus lack fine-granularity; 2) The brain activities (biological neurons, BNs) are limited to neural recordings of isolated cortical unit (i.e., voxel/region) and thus lack integrations and interactions among brain functions. To address those problems, in this study, we 1) define ANs with fine-granularity in transformer-based NLP models (BERT in this study) and measure their temporal activations to input text sequences; 2) define BNs as functional brain networks (FBNs) extracted from functional magnetic resonance imaging (fMRI) data to capture functional interactions in the brain; 3) couple ANs and BNs by maximizing the synchronization of their temporal activations. Our experimental results demonstrate 1) The activations of ANs and BNs are significantly synchronized; 2) the ANs carry meaningful linguistic/semantic information and anchor to their BN signatures; 3) the anchored BNs are interpretable in a neurolinguistic context. Overall, our study introduces a novel, general, and effective framework to link transformer-based NLP models and neural activities in response to language and may provide novel insights for future studies such as brain-inspired evaluation and development of NLP models.

cs.CL↗

arXiv2022

Using Targeted Maximum Likelihood Estimation to Estimate Treatment Effect with Longitudinal Continuous or Binary Data: A Systematic Evaluation of 28 Diabetes Clinical Trials

The primary analysis of clinical trials in diabetes therapeutic area often involves a mixed-model repeated measure (MMRM) approach to estimate the average treatment effect for longitudinal continuous outcome, and a generalized linear mixed model (GLMM) approach for longitudinal binary outcome. In this paper, we considered another estimator of the average treatment effect, called targeted maximum likelihood estimator (TMLE). This estimator can be a one-step alternative to model either continuous or binary outcome. We compared those estimators by simulation studies and by analyzing real data from 28 diabetes clinical trials. The simulations involved different missing data scenarios, and the real data sets covered a wide range of possible distributions of the outcome and covariates in real-life clinical trials for diabetes drugs with different mechanisms of action. For all the settings, adjusted estimators tended to be more efficient than the unadjusted one. In the setting of longitudinal continuous outcome, the MMRM approach with visits and baseline variables interaction appeared to dominate the performance of the MMRM considering the main effects only for the baseline variables while showing better or comparable efficiency to the TMLE estimator in both simulations and data applications. For modeling longitudinal binary outcome, TMLE generally outperformed GLMM in terms of relative efficiency, and its avoidance of the cumbersome covariance fitting procedure from GLMM makes TMLE a more advantageous estimator.

stat.AP↗

arXiv2022

Testing for Treatment Effect Twice Using Internal and External Controls in Clinical Trials

Leveraging external controls -- relevant individual patient data under control from external trials or real-world data -- has the potential to reduce the cost of randomized controlled trials (RCTs) while increasing the proportion of trial patients given access to novel treatments. However, due to lack of randomization, RCT patients and external controls may differ with respect to covariates that may or may not have been measured. Hence, after controlling for measured covariates, for instance by matching, testing for treatment effect using external controls may still be subject to unmeasured biases. In this paper, we propose a sensitivity analysis approach to quantify the magnitude of unmeasured bias that would be needed to alter the study conclusion that presumed no unmeasured biases are introduced by employing external controls. Whether leveraging external controls increases power or not depends on the interplay between sample sizes and the magnitude of treatment effect and unmeasured biases, which may be difficult to anticipate. This motivates a combined testing procedure that performs two highly correlated analyses, one with and one without external controls, with a small correction for multiple testing using the joint distribution of the two test statistics. The combined test provides a new method of sensitivity analysis designed for data fusion problems, which anchors at the unbiased analysis based on RCT only and spends a small proportion of the type I error to also test using the external controls. In this way, if leveraging external controls increases power, the power gain compared to the analysis based on RCT only can be substantial; if not, the power loss is small. The proposed method is evaluated in theory and power calculations, and applied to a real trial.

stat.ME↗

arXiv2022

HyperPrompt: Prompt-based Task-Conditioning of Transformers

Prompt-Tuning is a new paradigm for finetuning pre-trained language models in a parameter-efficient way. Here, we explore the use of HyperNetworks to generate hyper-prompts: we propose HyperPrompt, a novel architecture for prompt-based task-conditioning of self-attention in Transformers. The hyper-prompts are end-to-end learnable via generation by a HyperNetwork. HyperPrompt allows the network to learn task-specific feature maps where the hyper-prompts serve as task global memories for the queries to attend to, at the same time enabling flexible information sharing among tasks. We show that HyperPrompt is competitive against strong multi-task learning baselines with as few as $0.14\%$ of additional task-conditioning parameters, achieving great parameter and computational efficiency. Through extensive empirical experiments, we demonstrate that HyperPrompt can achieve superior performances over strong T5 multi-task learning baselines and parameter-efficient adapter variants including Prompt-Tuning and HyperFormer++ on Natural Language Understanding benchmarks of GLUE and SuperGLUE across many model sizes.

cs.CL↗

arXiv2022

A Unified and Biologically-Plausible Relational Graph Representation of Vision Transformers

Vision transformer (ViT) and its variants have achieved remarkable successes in various visual tasks. The key characteristic of these ViT models is to adopt different aggregation strategies of spatial patch information within the artificial neural networks (ANNs). However, there is still a key lack of unified representation of different ViT architectures for systematic understanding and assessment of model representation performance. Moreover, how those well-performing ViT ANNs are similar to real biological neural networks (BNNs) is largely unexplored. To answer these fundamental questions, we, for the first time, propose a unified and biologically-plausible relational graph representation of ViT models. Specifically, the proposed relational graph representation consists of two key sub-graphs: aggregation graph and affine graph. The former one considers ViT tokens as nodes and describes their spatial interaction, while the latter one regards network channels as nodes and reflects the information communication between channels. Using this unified relational graph representation, we found that: a) a sweet spot of the aggregation graph leads to ViTs with significantly improved predictive performance; b) the graph measures of clustering coefficient and average path length are two effective indicators of model prediction performance, especially when applying on the datasets with small samples; c) our findings are consistent across various ViT architectures and multiple datasets; d) the proposed relational graph representation of ViT has high similarity with real BNNs derived from brain science data. Overall, our work provides a novel unified and biologically-plausible paradigm for more interpretable and effective representation of ViT ANNs.

cs.NE↗

arXiv2022

Learning to Prompt for Open-Vocabulary Object Detection with Vision-Language Model

Recently, vision-language pre-training shows great potential in open-vocabulary object detection, where detectors trained on base classes are devised for detecting new classes. The class text embedding is firstly generated by feeding prompts to the text encoder of a pre-trained vision-language model. It is then used as the region classifier to supervise the training of a detector. The key element that leads to the success of this model is the proper prompt, which requires careful words tuning and ingenious design. To avoid laborious prompt engineering, there are some prompt representation learning methods being proposed for the image classification task, which however can only be sub-optimal solutions when applied to the detection task. In this paper, we introduce a novel method, detection prompt (DetPro), to learn continuous prompt representations for open-vocabulary object detection based on the pre-trained vision-language model. Different from the previous classification-oriented methods, DetPro has two highlights: 1) a background interpretation scheme to include the proposals in image background into the prompt training; 2) a context grading scheme to separate proposals in image foreground for tailored prompt training. We assemble DetPro with ViLD, a recent state-of-the-art open-world object detector, and conduct experiments on the LVIS as well as transfer learning on the Pascal VOC, COCO, Objects365 datasets. Experimental results show that our DetPro outperforms the baseline ViLD in all settings, e.g., +3.4 APbox and +3.0 APmask improvements on the novel classes of LVIS. Code and models are available at https://github.com/dyabel/detpro.

cs.CV↗

arXiv2022

A matching design for augmenting a randomized clinical trial with external control

The use of information from real world to assess the effectiveness of medical products is becoming increasingly popular and more acceptable by regulatory agencies. According to a strategic real-world evidence framework published by U.S. Food and Drug Administration, a hybrid randomized controlled trial that augments internal control arm with real-world data is a pragmatic approach worth more attention. In this paper, we aim to improve on existing matching designs for such a hybrid randomized controlled trial. In particular, we propose to match the entire concurrent randomized clinical trial (RCT) such that (1) the matched external control subjects used to augment the internal control arm are as comparable as possible to the RCT population, (2) every active treatment arm in an RCT with multiple treatments is compared with the same control group, and (3) matching can be conducted and the matched set locked before treatment unblinding to better maintain the data integrity. Besides a weighted estimator, we also introduce a bootstrap method to obtain its variance estimation. The finite sample performance of the proposed method is evaluated by simulations based on data from a real clinical trial.

stat.ME↗

arXiv2022

On perfectly matched layers of nonlocal wave equations in unbounded multi-scale media

A nonlocal perfectly matched layer (PML) is formulated for the nonlocal wave equation in the whole real axis and numerical discretization is designed for solving the reduced PML problem on a bounded domain. The nonlocal PML poses challenges not faced in PDEs. For example, there is no derivative in nonlocal models, which makes it impossible to replace derivates with complex ones. Here we provide a way of constructing the PML for nonlocal models, which decays the waves exponentially impinging in the layer and makes reflections at the truncated boundary very tiny. To numerically solve the nonlocal PML problem, we design the asymptotically compatible (AC) scheme for spatially nonlocal operator by combining Talbot's contour, and a Verlet-type scheme for time evolution. The accuracy and effectiveness of our approach are illustrated by various numerical examples.

math.NA↗

arXiv2022

LaMDA: Language Models for Dialog Applications

We present LaMDA: Language Models for Dialog Applications. LaMDA is a family of Transformer-based neural language models specialized for dialog, which have up to 137B parameters and are pre-trained on 1.56T words of public dialog data and web text. While model scaling alone can improve quality, it shows less improvements on safety and factual grounding. We demonstrate that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding. The first challenge, safety, involves ensuring that the model's responses are consistent with a set of human values, such as preventing harmful suggestions and unfair bias. We quantify safety using a metric based on an illustrative set of human values, and we find that filtering candidate responses using a LaMDA classifier fine-tuned with a small amount of crowdworker-annotated data offers a promising approach to improving model safety. The second challenge, factual grounding, involves enabling the model to consult external knowledge sources, such as an information retrieval system, a language translator, and a calculator. We quantify factuality using a groundedness metric, and we find that our approach enables the model to generate responses grounded in known sources, rather than responses that merely sound plausible. Finally, we explore the use of LaMDA in the domains of education and content recommendations, and analyze their helpfulness and role consistency.

cs.CL↗

arXiv2022

Charge Carrier Mediation and Ferromagnetism induced in MnBi6Te10 Magnetic Topological Insulators by antimony doping

A new kind of intrinsic magnetic topological insulators (MTI) MnBi2Te4 family have shed light on the observation of novel topological quantum effect such as quantum anomalous Hall effect (QAHE). However, the strong anti-ferromagnetic (AFM) coupling and high carrier concentration in the bulk hinder the practical applications. In closely related materials MnBi4Te7 and MnBi6Te10, the interlayer magnetic coupling is greatly suppressed by Bi2Te3 layer intercalation. However, AFM is still the ground state in these compounds. Here by magnetic and transport measurements, we demonstrate that Sb substitutional dopant plays a dual role in MnBi6Te10, which can not only adjust the charge carrier type and the concentration, but also induce the solid into a ferromagnetic (FM) ground state. AFM ground state region which is also close to the charge neutral point can be found in the phase diagram of Mn(SbxBi1-x)6Te10 when x ~ 0.25. An intrinsic FM-MTI candidate is thus demonstrated, and it may take a step further for the realization of high-quality and high-temperature QAHE and the related topological quantum effects in the future.

cond-mat.mtrl-sci↗

arXiv2021

Deep Personalized Glucose Level Forecasting Using Attention-based Recurrent Neural Networks

In this paper, we study the problem of blood glucose forecasting and provide a deep personalized solution. Predicting blood glucose level in people with diabetes has significant value because health complications of abnormal glucose level are serious, sometimes even leading to death. Therefore, having a model that can accurately and quickly warn patients of potential problems is essential. To develop a better deep model for blood glucose forecasting, we analyze the data and detect important patterns. These observations helped us to propose a method that has several key advantages over existing methods: 1- it learns a personalized model for each patient as well as a global model; 2- it uses an attention mechanism and extracted time features to better learn long-term dependencies in the data; 3- it introduces a new, robust training procedure for time series data. We empirically show the efficacy of our model on a real dataset.

cs.LG↗

arXiv2021

Analysis of an Incomplete Binary Outcome Dichotomized From an Underlying Continuous Variable in Clinical Trials

In many clinical trials, outcomes of interest include binary-valued endpoints. It is not uncommon that a binary-valued outcome is dichotomized from a continuous outcome at a threshold of clinical interest. To reach the objective, common approaches include (a) fitting the generalized linear mixed model (GLMM) to the dichotomized longitudinal binary outcome and (b) imputation method (MI): imputing the missing values in the continuous outcome, dichotomizing it into a binary outcome, and then fitting the generalized linear model for the "complete" data. We conducted comprehensive simulation studies to compare the performance of GLMM with MI for estimating risk difference and logarithm of odds ratio between two treatment arms at the end of study. In those simulation studies, we considered a range of multivariate distribution options for the continuous outcome (including a multivariate normal distribution, a multivariate t-distribution, a multivariate log-normal distribution, and the empirical distribution from a real clinical trial data) to evaluate the robustness of the estimators to various data-generating models. Simulation results demonstrate that both methods work well under those considered distribution options, but MI is more efficient with smaller mean squared errors compared to GLMM. We further applied both the GLMM and MI to 29 phase 3 diabetes clinical trials, and found that the MI method generally led to smaller variance estimates compared to GLMM.

stat.AP↗

arXiv2021

Online Ordering Platform City Distribution Based on Genetic Algorithm

Since the rising of the takeaway ordering platform, the M platform has taken the lead in the industry with its high-quality service. The increasing order volume leads the competition between platforms to reduce the distribution cost, which increases rapidly because of the unreasonable distribution route. By analyzing platform distribution's current situation, we study the vehicle routing problem of urban distribution on the M platform and minimize the distribution cost. Considering the constraints of the customer's expected delivery time and vehicle condition, we combine the different arrival times of the vehicle routing problem model using three soft time windows and solve the problem using a genetic algorithm (GA). The results show that our model and algorithm can design the vehicle path superior to the original model in terms of distribution cost and delivery time, thus providing decision support for the M platform to save distribution cost in urban distribution in the future.

stat.AP↗

arXiv2021

Numerical solution of a one-dimensional nonlocal Helmholtz equation by Perfectly Matched Layers

We consider the computation of a nonlocal Helmholtz equation by using Perfectly Matched Layer (PML). We first derive the nonlocal PML equation by extending PML modifications from the local operator to the nonlocal operator of integral form. We then give stability estimates of some weighted average value of the nonlocal Helmholtz solution and prove that (i) the weighted average value of the nonlocal PML solution decays exponentially in PML layers in one case; (ii) in the other case, the weighted average value of the nonlocal Helmholtz solution itself decays exponentially outside some domain. Particularly for a typical kernel function $γ_1(s)=\frac12 e^{-| s|}$, we obtain the Green's function of the nonlocal Helmholtz equation, and use the Green's function to further prove that (i) the nonlocal PML solution decays exponentially in PML layers in one case; (ii) in the other case, the nonlocal Helmholtz solution itself decays exponentially outside some domain. Based on our theoretical analysis, the truncated nonlocal problems are discussed and an asymptotic compatibility scheme is also introduced to solve the resulting truncated problems. Finally, numerical examples are provided to verify the effectiveness and validation of our nonlocal PML strategy and theoretical findings.

math.NA↗

arXiv2021

Perfectly Matched Layers for nonlocal Helmholtz equations II: multi-dimensional cases

Perfectly matched layers (PMLs) are formulated and applied to numerically solve nonlocal Helmholtz equations in one and two dimensions. In one dimension, we present the PML modifications for the nonlocal Helmholtz equation with general kernels and theoretically show its effectiveness in some sense. In two dimensions, we give the PML modifications in both Cartesian coordinates and polar coordinates. Based on the PML modifications, nonlocal Helmholtz equations are truncated in one and two dimensional spaces, and asymptotic compatibility schemes are introduced to discretize the resulting truncated problems. Finally, numerical examples are provided to study the "numerical reflections" by PMLs and demonstrate the effectiveness and validation of our nonlocal PML strategy.

math.NA↗