ML Techniques for High Throughput Analysis for PV Cells

Welcome to the repository documenting the work embarked upon during my journey as a Machine Learning Engineer Intern at CNRS, within the prestigious Institut Photovoltaïque d'Ile-de-France (IPVF). This repository showcases the research and development efforts under the guidance of esteemed mentors Philip Schulz, Jean-Baptiste Puel, and Arthur Julien, as part of the captivating project titled "Analysis of Interfaces in Perovskite-Based Tandem Solar Cells with Advanced Spectroscopy and Data Science Techniques."

Project Overview

This project focuses on leveraging advanced machine learning techniques to analyze interfaces in perovskite-based tandem solar cells. The main tasks undertaken in this research include:

1. Introducing Uncertainties for Bayesian Convolutional Neural Networks
   Implementing Bayesian Convolutional Neural Networks (BCNNs) to account for uncertainties in predictions, enhancing the robustness and interpretability of the model.

2. Incorporating the Bayesian Model with SparseDenseNet Network
   Integrating the Bayesian model with a SparseDenseNet architecture to optimize performance in handling the complex data associated with tandem solar cells.

3. Hyperparameter Tuning of Neural Networks
   Systematically tuning the hyperparameters of the neural networks to achieve optimal performance, balancing the trade-offs between various metrics.

4. Error Function Comparison: MSE vs. MAE
   Evaluating the performance of the model using different error functions, specifically Mean Squared Error (MSE) and Mean Absolute Error (MAE), to determine the most suitable metric for this analysis.

Repository Contents

• Code: The scripts and Jupyter notebooks implementing the BCNN, SparseDenseNet, and other neural network models.
• Documentation: Detailed explanations and methodologies used throughout the project.
• Reference Notebooks: Supporting notebooks used for data analysis, model training, and validation.
• Results: Outputs, visualizations, and performance metrics from the experiments conducted.

Getting Started

To get started, clone this repository and explore the notebooks provided. Ensure that the required dependencies are installed as listed in the requirements.txt.

git clone https://github.com/yourusername/CNN_Peak_Fitting_TRPES_Spectra_Analysis.git
cd CNN_Peak_Fitting_TRPES_Spectra_Analysis
pip install -r requirements.txt

Contributions

This work has been carried out with the support of IPVF and under the mentorship of Philip Schulz, Jean-Baptiste Puel, and Arthur Julien. Their guidance has been instrumental in the progress of this project.

---