Dimensionality Reduction for Studying Diffuse Circumgalactic Medium.
Implemented dimensionality reduction techniques to (best) preserve classification accuracy on reduce feature space.
Used Deep Autoencoders for dimensionality reduction; and DNN, XGBoost for classification. Results and procedures described in Technical Report.
├── LICENSE
├── Part_1
│ ├── Solution_Part_1.md
│ └── Solution_Part_1.pdf
├── Part_2
│ ├── SpectrumCalc.ipynb
│ └── SpectrumCalc.pdf
├── Part_3
│ ├── CGM_Technical_Report.pdf
│ ├── ColabNotebooks
│ │ ├── DAE_DNN.ipynb
│ │ ├── DAE.ipynb
│ │ ├── DAE_XGB.ipynb
│ │ ├── VanillaDNN.ipynb
│ │ └── VanillaXGB.ipynb
│ └── NotebookOutputsPDF
│ ├── DAE_DNN.pdf
│ ├── DAE.pdf
│ ├── DAE_XGB.pdf
│ ├── VanillaDNN.pdf
│ └── VanillaXGB.pdf
└── README.md
$ git clone https://github.com/ShivenTripathi/Dimensionality-Reduction-CGM.git
- Run code cells in SpectrumCalc.ipynb, to generate spectrum outputs
-
Create a shortcut of Drive Folder, containing dataset, models and encodings, to your drive
-
Run the Colab Notebooks after changing the user directory to your Drive.