arXiv subjects

Tien Dat Nguyen

Publications and source records attributed to Tien Dat Nguyen.

13 recordsLinked to original sources

arXiv2026

CoDiR: Confidence-Guided Diffusion Refinement for Semi-Supervised Histopathology Segmentation

Semi-supervised histopathology segmentation is challenging due to scarce annotations and unreliable pseudo-labels in ambiguous gland regions. To address this problem, we propose Confidence-Guided Diffusion Refinement (CoDiR), a semi-supervised framework that combines a Mean Teacher segmentation model with diffusion-based pseudo-label refinement. Given an unlabeled image, the teacher first produces a soft prediction, and only low-confidence regions are refined by a conditional diffusion model trained to capture plausible mask structures from labeled data. The refined mask is then fused with reliable teacher predictions and used to train the student with confidence weighting and consistency regularization. On the GlaS and CRAG datasets CoDiR reaches 88.09\% and 89.83\% mDice with 10\% labeled data, and 89.19\% and 90.29\% mDice with 20\%, matching or exceeding the strongest published method on seven of the eight benchmark metrics. Ablations attribute the largest single contribution to the refinement module, which adds +6.36\% mDice over the Mean Teacher baseline. The implementation code is publicly available at: https://github.com/vongla345/codir

cs.CV↗

arXiv2026

Towards Pedagogically Aligned LLM Tutors for Math Mistake Remediation

Large language models have strong potential for use in intelligent tutoring systems, but they often fail to follow effective pedagogical strategies, such as guiding students without revealing final answers. We study the application of a two-stage alignment pipeline for math mistake remediation, combining supervised fine-tuning on tutoring dialogs with Direct Preference Optimization on synthetic preference pairs. We construct a dataset that integrates existing tutoring corpora with synthetic data generated along pedagogical dimensions, such as scaffolding and factuality, and study different input configurations that incorporate solution correctness and gold answers. Experiments show that this approach improves both factual accuracy and pedagogical quality over base models and existing tutoring models. Human evaluation further indicates that our best model is competitive with a strong proprietary baseline, while providing additional benefits in terms of openness, transparency, and reproducibility. Our results highlight the effectiveness of preference-based pedagogical alignment, while also revealing challenges in reliably evaluating tutoring quality.

cs.CL↗

arXiv2026

UniSemAlign: Text-Prototype Alignment with a Foundation Encoder for Semi-Supervised Histopathology Segmentation

Semi-supervised semantic segmentation in computational pathology remains challenging due to scarce pixel-level annotations and unreliable pseudo-label supervision. We propose UniSemAlign, a dual-modal semantic alignment framework that enhances visual segmentation by injecting explicit class-level structure into pixel-wise learning. Built upon a pathology-pretrained Transformer encoder, UniSemAlign introduces complementary prototype-level and text-level alignment branches in a shared embedding space, providing structured guidance that reduces class ambiguity and stabilizes pseudo-label refinement. The aligned representations are fused with visual predictions to generate more reliable supervision for unlabeled histopathology images. The framework is trained end-to-end with supervised segmentation, cross-view consistency, and cross-modal alignment objectives. Extensive experiments on the GlaS and CRAG datasets demonstrate that UniSemAlign substantially outperforms recent semi-supervised baselines under limited supervision, achieving Dice improvements of up to 2.6% on GlaS and 8.6% on CRAG with only 10% labeled data, and strong improvements at 20% supervision. Code is available at: https://github.com/thailevann/UniSemAlign

cs.CV↗

arXiv2026

High-dimensional Many-to-many-to-many Mediation Analysis

We study high-dimensional mediation analysis in which exposures, mediators, and outcomes are all multivariate, and both exposures and mediators may be high-dimensional. We formalize this as a many (exposures)-to-many (mediators)-to-many (outcomes) (MMM) mediation analysis problem. Methodologically, MMM mediation analysis simultaneously performs variable selection for high-dimensional exposures and mediators, estimates the indirect effect matrix (i.e., the coefficient matrices linking exposure-to-mediator and mediator-to-outcome pathways), and enables prediction of multivariate outcomes. Theoretically, we show that the estimated indirect effect matrices are consistent and element-wise asymptotically normal, and we derive error bounds for the estimators. To evaluate the efficacy of the MMM mediation framework, we first investigate its finite-sample performance, including convergence properties, the behavior of the asymptotic approximations, and robustness to noise, via simulation studies. We then apply MMM mediation analysis to data from the Alzheimer's Disease Neuroimaging Initiative to study how cortical thickness of 202 brain regions may mediate the effects of 688 genome-wide significant single nucleotide polymorphisms (SNPs) (selected from approximately 1.5 million SNPs) on eleven cognitive-behavioral and diagnostic outcomes. The MMM mediation framework identifies biologically interpretable, many-to-many-to-many genetic-neural-cognitive pathways and improves downstream out-of-sample classification and prediction performance. Taken together, our results demonstrate the potential of MMM mediation analysis and highlight the value of statistical methodology for investigating complex, high-dimensional multi-layer pathways in science. The MMM package is available at https://github.com/THELabTop/MMM-Mediation.

stat.ME↗

arXiv2025

Adaptive Density Estimation Using Projection Kernels and Penalized Comparison to Overfitting

In this work, we study wavelet projection estimators for density estimation, focusing on their construction from $\mathcal{S}$-regular, compactly supported wavelet bases. A key aspect of such estimators is the choice of the resolution level, which controls the balance between bias and variance. To address this, we employ the Penalized Comparison to Overfitting (PCO) method, providing a fully data-driven selection of the multiresolution level. This approach ensures both statistical accuracy and computational feasibility, while retaining the adaptability of wavelet methods. Our results establish an oracle inequality for the proposed estimator and show that it attains the optimal rate of convergence in density estimation.

math.ST↗

arXiv2025

Adaptive estimation for nonparametric circular regression with errors in variables

This paper investigates the nonparametric estimation of a circular regression function in an errors-in-variables framework. Two settings are studied, depending on whether the covariates are circular or linear. Adaptive estimators are constructed and their theoretical performance is assessed through convergence rates over Sobolev and H\"older smoothness classes. Numerical experiments on simulated and real datasets illustrate the practical relevance of the methodology.

math.ST↗

arXiv2025

Enhancing Retrieval Augmented Generation with Hierarchical Text Segmentation Chunking

Retrieval-Augmented Generation (RAG) systems commonly use chunking strategies for retrieval, which enhance large language models (LLMs) by enabling them to access external knowledge, ensuring that the retrieved information is up-to-date and domain-specific. However, traditional methods often fail to create chunks that capture sufficient semantic meaning, as they do not account for the underlying textual structure. This paper proposes a novel framework that enhances RAG by integrating hierarchical text segmentation and clustering to generate more meaningful and semantically coherent chunks. During inference, the framework retrieves information by leveraging both segment-level and cluster-level vector representations, thereby increasing the likelihood of retrieving more precise and contextually relevant information. Evaluations on the NarrativeQA, QuALITY, and QASPER datasets indicate that the proposed method achieved improved results compared to traditional chunking techniques.

cs.CL↗

arXiv2023

Learning Symmetrization for Equivariance with Orbit Distance Minimization

We present a general framework for symmetrizing an arbitrary neural-network architecture and making it equivariant with respect to a given group. We build upon the proposals of Kim et al. (2023); Kaba et al. (2023) for symmetrization, and improve them by replacing their conversion of neural features into group representations, with an optimization whose loss intuitively measures the distance between group orbits. This change makes our approach applicable to a broader range of matrix groups, such as the Lorentz group O(1, 3), than these two proposals. We experimentally show our method's competitiveness on the SO(2) image classification task, and also its increased generality on the task with O(1, 3). Our implementation will be made accessible at https://github.com/tiendatnguyen-vision/Orbit-symmetrize.

cs.LG↗

arXiv2023

Learning Probabilistic Symmetrization for Architecture Agnostic Equivariance

We present a novel framework to overcome the limitations of equivariant architectures in learning functions with group symmetries. In contrary to equivariant architectures, we use an arbitrary base model such as an MLP or a transformer and symmetrize it to be equivariant to the given group by employing a small equivariant network that parameterizes the probabilistic distribution underlying the symmetrization. The distribution is end-to-end trained with the base model which can maximize performance while reducing sample complexity of symmetrization. We show that this approach ensures not only equivariance to given group but also universal approximation capability in expectation. We implement our method on various base models, including patch-based transformers that can be initialized from pretrained vision transformers, and test them for a wide range of symmetry groups including permutation and Euclidean groups and their combinations. Empirical tests show competitive results against tailored equivariant architectures, suggesting the potential for learning equivariant functions for diverse groups using a non-equivariant universal base architecture. We further show evidence of enhanced learning in symmetric modalities, like graphs, when pretrained from non-symmetric modalities, like vision. Code is available at https://github.com/jw9730/lps.

cs.LG↗

arXiv2022

Pure Transformers are Powerful Graph Learners

We show that standard Transformers without graph-specific modifications can lead to promising results in graph learning both in theory and practice. Given a graph, we simply treat all nodes and edges as independent tokens, augment them with token embeddings, and feed them to a Transformer. With an appropriate choice of token embeddings, we prove that this approach is theoretically at least as expressive as an invariant graph network (2-IGN) composed of equivariant linear layers, which is already more expressive than all message-passing Graph Neural Networks (GNN). When trained on a large-scale graph dataset (PCQM4Mv2), our method coined Tokenized Graph Transformer (TokenGT) achieves significantly better results compared to GNN baselines and competitive results compared to Transformer variants with sophisticated graph-specific inductive bias. Our implementation is available at https://github.com/jw9730/tokengt.

cs.LG↗

arXiv2022

Adaptive warped kernel estimation for nonparametric regression with circular responses

In this paper, we deal with nonparametric regression for circular data, meaning that observations are represented by points lying on the unit circle. We propose a kernel estimation procedure with data-driven selection of the bandwidth parameter. For this purpose, we use a warping strategy combined with a Goldenshluger-Lepski type estimator. To study optimality of our methodology, we consider the minimax setting and prove, by establishing upper and lower bounds, that our procedure is nearly optimal on anisotropic Holder classes of functions for pointwise estimation. The obtained rates also reveal the specific nature of regression for circular responses. Finally, a numerical study is conducted, illustrating the good performances of our approach.

math.ST↗

arXiv2020

Statistical deconvolution of the free Fokker-Planck equation at fixed time

We are interested in reconstructing the initial condition of a non-linear partial differential equation (PDE), namely the Fokker-Planck equation, from the observation of a Dyson Brownian motion at a given time $t>0$. The Fokker-Planck equation describes the evolution of electrostatic repulsive particle systems, and can be seen as the large particle limit of correctly renormalized Dyson Brownian motions. The solution of the Fokker-Planck equation can be written as the free convolution of the initial condition and the semi-circular distribution. We propose a nonparametric estimator for the initial condition obtained by performing the free deconvolution via the subordination functions method. This statistical estimator is original as it involves the resolution of a fixed point equation, and a classical deconvolution by a Cauchy distribution. This is due to the fact that, in free probability, the analogue of the Fourier transform is the R-transform, related to the Cauchy transform. In past literature, there has been a focus on the estimation of the initial conditions of linear PDEs such as the heat equation, but to the best of our knowledge, this is the first time that the problem is tackled for a non-linear PDE. The convergence of the estimator is proved and the integrated mean square error is computed, providing rates of convergence similar to the ones known for non-parametric deconvolution methods. Finally, a simulation study illustrates the good performances of our estimator.

math.PR↗

arXiv2020

bertha: Project Skeleton for Scientific Software

Science depends heavily on reliable and easy-to-use software packages, such as mathematical libraries or data analysis tools. Developing such packages requires a lot of effort, which is too often avoided due to the lack of funding or recognition. In order to reduce the efforts required to create sustainable software packages, we present a project skeleton that ensures the best software engineering practices from the start of a project, or serves as reference for existing projects.

cs.MS↗