A5100764539- Stochastic Gradient Optimization Techniques
- Privacy-Preserving Technologies in Data
- Cryptography and Data Security
- Data Management and Algorithms
- Coconut Research and Applications
- Conducting polymers and applications
- Trace Elements in Health
- Data Mining Algorithms and Applications
- Optimization and Variational Analysis
- Blockchain Technology Applications and Security
- Advanced Statistical Process Monitoring
- Adversarial Robustness in Machine Learning
- Cryptography and Residue Arithmetic
- Agricultural and Environmental Management
- Viral Infections and Immunology Research
- SARS-CoV-2 and COVID-19 Research
- Integrated Circuits and Semiconductor Failure Analysis
- Advanced Database Systems and Queries
- Complexity and Algorithms in Graphs
- Gas Sensing Nanomaterials and Sensors
- Optimization and Search Problems
- Aerospace Engineering and Control Systems
- Full-Duplex Wireless Communications
- Distributed Sensor Networks and Detection Algorithms
- Industrial Vision Systems and Defect Detection
Vietnam National University Ho Chi Minh City
2024
Ho Chi Minh City University of Technology
2024
University of Connecticut
2021
Can Tho University
2020
Hanoi University of Science and Technology
2014-2019
Vietnam Academy of Science and Technology
2013
Lattice Semiconductor (United States)
2013
The feasibility of federated learning is highly constrained by the server-clients infrastructure in terms network communication. Most newly launched smartphones and IoT devices are equipped with GPUs or sufficient computing hardware to run powerful AI models. However, case original synchronous learning, client suffer waiting times regular communication between clients server required. This implies more sensitivity local model training irregular missed updates, hence, less limited scalability...
In this paper, a high-efficiency voltage doubler rectifier for radio frequency identification (RFID) applications is presented. The study focuses on new internal V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Th</sub> cancellation (IVC) technique which combination between conventional IVC circuit and proposed switches to improve power conversion efficiency (PCE) (VCE). Proposed designed using 0.13um CMOS technology passive tags of high (HF)...
A yield analysis method using basic and inline defect information to statistically determine significant root-causes of loss in semiconductor manufacturing is presented. Using simple statistics on pass/fail die maps, the goal this provide fab process line with feedback impact for inspected layers. Quantifying these losses, terms percent statistical confidence allows set priorities reduction work achieve maximum enhancement. statistics, valid sources can be separated from nonkiller or...
<p>The national institute of standards and technology (NIST) has presented its draft the module-lattice-based key-encapsulation mechanism standard (MLBKEMS), choosing cryptographic suite for algebraic lattices (CRYSTALS)- Kyber as base encryption. Existing hardware implementations modern cryptography will need to process new efficiently. The primary in CRYSTALS-Kyber (KEM) is number theoretic transform (NTT), which requires heavy computing power. This paper contributes an efficient...
Hogwild! implements asynchronous Stochastic Gradient Descent (SGD) where multiple threads in parallel access a common repository containing training data, perform SGD iterations and update shared state that represents jointly learned (global) model. We consider big data analysis is distributed among local sets heterogeneous way -- we wish to move computations compute nodes resides. The results of these are aggregated by central "aggregator" which mimics Hogwild!. show how can start choosing...
Second-order necessary optimality conditions for an optimal control problem with a nonconvex cost function and state-control constraints are studied in this paper. By establishing abstract result on second-order mathematical programming problem, we obtain problem.
Each round in Differential Private Stochastic Gradient Descent (DPSGD) transmits a sum of clipped gradients obfuscated with Gaussian noise to central server which uses this update global model often represents deep neural network. Since the are computed separately, we call Individual Clipping (IC), networks like resnet-18 cannot use Batch Normalization Layers (BNL) is crucial component for achieving high accuracy. To utilize BNL, introduce (BC) where, instead clipping single as orginal...
This study analyses the normal Sap coconut and embryo cultured models as priority subjects on researching crop varieties adapted to drought. The data used were a large number method with of in Tra Vinh province. There was no difference between two regard average roughage time 3.6 years. stable age ranged from 5.6 5.9 years after growing an yield 47 fruits per tree yearly. investment cost for model is 2.5 times higher than that conventional , while profit 3.9 coconut. closely related mutant...
Classical differential private DP-SGD implements individual clipping with random subsampling, which forces a mini-batch SGD approach. We provide general algorithmic framework that goes beyond and allows any possible first order optimizers (e.g., classical momentum based approaches) in combination batch clipping, clips an aggregate of computed gradients rather than summing clipped (as is done clipping). The also admits sampling techniques subsampling such as shuffling. Our DP analysis follows...
We introduce a multiple target optimization framework for DP-SGD referred to as pro-active DP. In contrast traditional DP accountants, which are used track the expenditure of privacy budgets, scheme allows one {\it a-priori} select parameters based on fixed budget (in terms $\epsilon$ and $\delta$) in such way optimize anticipated utility (test accuracy) most. To achieve this objective, we first propose significant improvements moment account method, presenting closed-form...