A5051145133- Photonic and Optical Devices
- Advanced Fiber Optic Sensors
- Semiconductor Lasers and Optical Devices
- Advanced Fiber Laser Technologies
- Building Energy and Comfort Optimization
- Machine Learning and Data Classification
- Machine Learning and Algorithms
- Adsorption and Cooling Systems
- Energy Efficiency and Management
- Advanced Surface Polishing Techniques
- Neural Networks and Applications
- Advanced Multi-Objective Optimization Algorithms
- Gaussian Processes and Bayesian Inference
- Optical Systems and Laser Technology
- Heat Transfer and Optimization
- Optical Coatings and Gratings
- Geotechnical Engineering and Underground Structures
- Neural Networks and Reservoir Computing
- Vibration and Dynamic Analysis
- Photonic Crystal and Fiber Optics
Ferdinand-Braun-Institut
2023-2024
Amirkabir University of Technology
2021
Deep learning models, with a prerequisite of large databases, are common approaches in applying machine for inverse design photonics. For these less expensive, approximate methods usually used to generate which limit their applications. In this study, we compare the performance data-efficient (ML) models predicting characteristics surface Bragg gratings semiconductor ridge waveguides. We employ 3D finite-difference time-domain method is very accurate but time-consuming database. analyze...
Abstract Acquiring a substantial number of data points for training accurate machine learning (ML) models is big challenge in scientific fields where collection resource-intensive. Here, we propose novel approach constructing minimal yet highly informative database ML complex multi-dimensional parameter spaces. To achieve this, mimic the underlying relation between output and input parameters using Gaussian process regression (GPR). Using set known data, GPR provides predictive means...
This study focuses on accurately fit of the main and side lobes reflectance obtained through precise 3D FDTD simulations using coupled-mode-theory. approach based surrogate modeling reduces reliance time-consuming simulations.
Here we introduce efficient machine learning models trained on a 3D FDTD simulation-based database to predict Bragg grating characteristics from the main and side lobes of reflectance spectra fitted by coupled mode theory.
Abstract This study discusses the importance of accurately calculating optical response Bragg gratings and challenges associated with 3D finite-difference time-domain (FDTD) method for simulating large-scale structures. The grating section in monolithic extended cavity diode lasers is substantial size, making FDTD simulations computationally challenging due to their complexity. In order assess accuracy simulations, we compare them experimental results. Using a precise model design, involving...
Bragg gratings are an essential component of semiconductor lasers. One the most precise methods to calculate optical response such a is 3D FDTD method. However, due its computational effort, it usually not used for simulation large structures. Here, we investigate performance 2D and Finite-Difference Time-Domain (FDTD) grating simulations. We demonstrate, that method can be structures interest, while only short up 500 μm. show results utilized set efficient simulation. demonstrate...
In this paper, we demonstrate a possibility to predict the characteristics of semiconductor-based Bragg gratings using machine learning methods. We perform 2D simulations and calculate reflectance create database. With obtained data, train ML models shape upper part main peak reflectance. compare performance widely used neural network with various different on our data high accuracy optimized XGBoost method.
Precisely stabilizing laser frequency is crucial for advancing technology and unlocking the full potential of various quantum technologies. Here, we propose a compact device semiconductor through mode coupling effects, which provides enhanced sensitivity. Our proposed architecture features main ridge waveguide with Bragg grating, flanked by two curved waveguides. This configuration exhibits an optical phenomenon characterized transmission crossing at wavelength grating. Using particle swarm...
The present study investigated a counter-flow cooling tower performance by integrating the Artificial Neural Networks and Intelligent Optimisation Algorithms (ANN-IOAs). For this purpose, two scenarios were evaluated. In first scenario, inlet air wet-bulb temperature (Taw), dry bulb (Tad), water to mass flow rate ratio (mw/ma), rotor speed (υ) input parameters for ANNs, while output (Two) was ANNs output. second same applied scenario used as variables efficiency (ε) considered an parameter....
Acquiring a substantial number of data points for training accurate machine learning (ML) models is big challenge in scientific fields where collection resource-intensive. Here, we propose novel approach constructing minimal yet highly informative database ML complex multi-dimensional parameter spaces. To achieve this, mimic the underlying relation between output and input parameters using Gaussian process regression (GPR). Using set known data, GPR provides predictive means standard...