A5090184417- Topological and Geometric Data Analysis
- Generative Adversarial Networks and Image Synthesis
- Face and Expression Recognition
- Microplastics and Plastic Pollution
- Anomaly Detection Techniques and Applications
- 3D Shape Modeling and Analysis
- Adversarial Robustness in Machine Learning
- Stochastic Gradient Optimization Techniques
- Recycling and Waste Management Techniques
- Image Retrieval and Classification Techniques
- Computer Graphics and Visualization Techniques
- Asphalt Pavement Performance Evaluation
- Machine Learning and Algorithms
- Gaussian Processes and Bayesian Inference
- Markov Chains and Monte Carlo Methods
- Domain Adaptation and Few-Shot Learning
- Advanced Statistical Methods and Models
- Advanced Image and Video Retrieval Techniques
- Groundwater flow and contamination studies
- Leprosy Research and Treatment
- Model Reduction and Neural Networks
- Advanced Neuroimaging Techniques and Applications
- Sparse and Compressive Sensing Techniques
- Advanced Bandit Algorithms Research
- Digital Image Processing Techniques
Toulouse School of Economics
2021-2024
Université Grenoble Alpes
2024
Laboratoire Jean Kuntzmann
2024
Ho Chi Minh City University of Technology
2021-2023
Vietnam National University Ho Chi Minh City
2010-2023
Université de Toulouse
2021-2023
The Institute of Statistical Mathematics
2023
RIKEN Center for Advanced Intelligence Project
2019-2021
National Institute for Materials Science
2018-2020
Cat Tien National Park
2020
Wastewater-treatment plants (WWTPs) are considered significant point sources of microplastics (MiP) in the receiving waters; MiP release is poorly estimated developing countries. abundance, recovery efficiency and daily load to waters were explored treatment stage facilities four WWTPs Vietnam, located Ho Chi Minh City, Thu Dau Mot, Da Lat. abundance varied from 1860 items m−3 125,000 influents between 140 813 final effluents. The MiP-removal was highest Lat wastewater-treatment plant (DL...
Algebraic topology methods have recently played an important role for statistical analysis with complicated geometric structured data such as shapes, linked twist maps, and material data. Among them, \textit{persistent homology} is a well-known tool to extract robust topological features, outputs \textit{persistence diagrams} (PDs). However, PDs are point multi-sets which can not be used in machine learning algorithms vector To deal it, emerged approach use kernel methods, appropriate...
Learning an effective representation of 3D point clouds requires a good metric to measure the discrepancy between two sets, which is non-trivial due their irregularity. Most previous works resort using Chamfer or Earth Mover’s distance, but those metrics are either ineffective in measuring differences computationally expensive. In this paper, we conduct systematic study with extensive experiments on distance for clouds. From study, propose use sliced Wasserstein and its variants learning...
Moving beyond $L^p$ geometric structure, Orlicz-Wasserstein (OW) leverages a specific class of convex functions for Orlicz structure. While OW remarkably helps to advance certain machine learning approaches, it has high computational complexity due its two-level optimization formula. Recently, Le et al. (2024) exploits graph structure propose generalized Sobolev transport (GST), i.e., scalable variant OW. However, GST assumes that input measures have the same mass. Unlike optimal (OT), is...
We investigate the Sobolev IPM problem for probability measures supported on a graph metric space. is an important instance of integral metrics (IPM), and obtained by constraining critic function within unit ball defined norm. In particular, it has been used to compare crucial several theoretical works in machine learning. However, our knowledge, there are no efficient algorithmic approaches compute effectively, which hinders its practical applications. this work, we establish relation...
This research comprehensively assessed the water quality in three urban canals within Ho Chi Minh City (HCMC), Vietnam, representing waterways affected by various pollutants. In rapid pace of urbanization and industrialization has exerted significant pressure on bodies, turning many into repositories for untreated domestic sewage, industrial discharge, runoff. City, as nation's largest economic hub, been particularly affected, with serving both drainage systems informal waste disposal sites....
Increasing attention has been paid to materials informatics approaches that promise efficient and fast discovery optimization of functional inorganic materials. Technical breakthrough is urgently requested advance this field efforts have made in the development descriptors encode or represent characteristics crystalline solids, such as chemical composition, crystal structure, electronic etc. We propose a general representation scheme for solids lifts restrictions on atom ordering, cell...
Optimal transport (\OT) theory defines a powerful set of tools to compare probability distributions. \OT~suffers however from few drawbacks, computational and statistical, which have encouraged the proposal several regularized variants OT in recent literature, one most notable being \textit{sliced} formulation, exploits closed-form formula between univariate distributions by projecting high-dimensional measures onto random lines. We consider this work more general family ground metrics,...
In view of training increasingly complex learning architectures, we establish a nonsmooth implicit function theorem with an operational calculus. Our result applies to most practical problems (i.e., definable problems) provided that form the classical invertibility condition is fulfilled. This approach allows for formal subdifferentiation: instance, replacing derivatives by Clarke Jacobians in usual differentiation formulas fully justified wide class problems. Moreover this calculus entirely...
Abstract Natural forests in Vietnam have experienced rapid declines the last 70 years, as a result of degradation from logging and conversion natural to timber rubber plantations. Degradation leads loss biodiversity ecosystem services, impacting livelihoods surrounding communities. Efforts address ongoing forests, through mechanisms such Reduced Emissions Deforestation (REDD+), require an understanding links between forest local communities, which rarely been studied Vietnam. We combined...
A combination of a submerged membrane filtration system and powdered activated carbon (PAC) was investigated for nonylphenol ethoxylates removal. Both flux initial dosage had significant effects on the micropollutants removal efficiency. The best performance achieved under 20 L/m2.h 50 mg/L. efficiencies obtained at 75±5% in first 60 hours, then decreased 55±7% 23±11% following respectively. As observed, over 65% dissolved organic mass adsorbed into that suspended bulk phase, remainder...
Neural architecture search (NAS) automates the design of deep neural networks. One main challenges in searching complex and non-continuous architectures is to compare similarity networks that conventional Euclidean metric may fail capture. Optimal transport (OT) resilient such structure by considering minimal cost for transporting a network into another. However, OT generally not negative definite which limit its ability build positive-definite kernels required many kernel-dependent...
.Risk minimization for nonsmooth nonconvex problems naturally leads to first-order sampling or, by an abuse of terminology, stochastic subgradient descent. We establish the convergence this method in path-differentiable case and describe more precise results under additional geometric assumptions. recover improve from Ermoliev Norkin [Cybern. Syst. Anal., 34 (1998), pp. 196–215] using a different approach: conservative calculus ODE method. In definable case, we show that avoids artificial...