haitaozhao / Neural-Component-Analysis

This repository is mainly to show the source code of neural component analysis.



 Neural Component Analysis for Fault Detection

---

• Code for the paper "Neural Component Analysis for Fault Detection"
• Directory "data" contains the simulated TE data for experiments
• File "myfunction_pca_kde.m" is designed for performing PCA on TE data
• File "myfunction_kernel_pca_kde.m" is designed for performing Kernel PCA on TE data
• File "myfunction_kernel_pca_kde.m" needs KPCA code which can be found in the Github link of Deng Cai
• Please run the matlab code with the data in the same directory
• File "kde.m" is used for density estimation
• File "autoencoder.py" is designed for performing autoencoder on TE data
• File "nca.py" is designed for performing NCA on TE data
• The python codes is developed with the package PyTorch. Please install the following python libraries before running the python codes

---

Requirements for running the python codes "nca.py" and "autoencoder.py"

Please install the following python libraries before running the code

• python==3.52
• numpy==1.13.3
• PyTorch==0.20
• scikit-learn==0.19.0

For the GPU acceleration with PyTorch, please refer to PyTorch

Run

On linux:

• python3 nca.py
• python3 autoencoder.py

---

Citation Information:

Haitao Zhao, Neural component analysis for fault detection, Chemometrics and Intelligent Laboratory Systems, Volume 176, 2018, pp. 11-21.