DimRed

Dimensionality Reduction

The folder is structured as follows:

• helpers.py: python file containing all the helper functions.

• tests.ipynb: jupyter notebook containing some tests to check the normalization.

• exploration.ipynb: jupyter notebook exploring the thickness data.

• Stacked_Data.ipynb: jupyter notebook exploring the stacked data, plotting the distribution of the thickness layers, and applying one model as test for the first time.

• Dim_Red_ModelComparisons.ipynb: jupyter notebook running a serie of dimensionality reduction models on the thickness layers in order to compare the results.

  4 Model Comparisons on Thickness Layers

  • 4.1 One model study
    • 4.1.1 NMF model
    • 4.1.2 t-SNE
    • 4.1.3 UMAP
  • 4.2 compare supervised dim. reduction models

• Dim_Red_GradientImage_ModelComparisons.ipynb: jupyter notebook running a serie of dimensionality reduction models on the gradient of the OCT images in order to compare the results.

  5 Gradient Image: Model Comparisons

  • 5.1 Preprocess the OCTs
  • 5.2 One model study
    • 5.2.1 t-SNE
    • 5.2.2 UMAP
  • 5.3 compare supervised dim. reduction models

• HorData.ipynb: jupyter notebook considering the vectors as horizontal. Performs dimensionality reduction on the horizontal data, i.e. data with shape $(14\cdot 18 \cdot 5, 768)$. Table of contents:

  6 Horizontal Data

  • 6.1: TSNE
    • 6.1.1: TSNE with different parameters
  • 6.2: UMAP
    • 6.2.1: UMAP with different parameters
  • 6.3: Layer-labeled horizontal data
    • 6.3.1: TSNE with layer labels
    • 6.3.2: UMAP with layer labels
  • 6.4 Dim. Red. on single layers
    • 6.4.1: TSNE on single layers
    • 6.4.2: UMAP on single layers

To summarize, the most important files are Dim_Red_ModelComparisons.ipynb, Dim_Red_GradientImage_ModelComparisons.ipynb and HorData.ipynb, which compare models on thickness layers, gradient filtered scans and horizontal data, thanks to the helper functions in helpers.py. The other files are for data exploration and testing.