shiventripathi / dimensionality-reduction-cgm

Dimensionality Reduction for Studying Diffuse Circumgalactic Medium

Dimensionality-Reduction-CGM

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.

TLDR:

Part 1

Solution_Part_1.pdf

Part 2

SpectrumCalc.pdf

Part 3

CGM_Technical_Report.pdf

Repository Structure

├── 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

Usage

$ git clone https://github.com/ShivenTripathi/Dimensionality-Reduction-CGM.git

Part 2:

• Run code cells in SpectrumCalc.ipynb, to generate spectrum outputs

Part 3:

1. Create a shortcut of Drive Folder, containing dataset, models and encodings, to your drive

2. Run the Colab Notebooks after changing the user directory to your Drive.