/SIE-Project

SIE project for improving uncertainty measures in CNN using Density Forests

Primary LanguageJupyter NotebookMIT LicenseMIT

SIE Project 2017

Folder for various codes for the SIE project at EPFL in Fall Semester 2017.

Files

Code/

Visualizations

  • decision_tree.ipynb: Decision Trees and Random Forest on randomly generated labelled data
  • density_tree.ipynb: Density Trees on randomly generated labelled data
  • MNIST.ipynb: Trainig of a CNN on the MNIST dataset, retrieval of the FC layer activation weights, Density Forest

Code/density_tree

Package for implementation of Decision Trees, Density Forests and Random Forests

  • create_data.py: functions for generating labelled and unlabelled data
  • decision_tree.py: data structure for decision tree nodes
  • decision_tree_create.py: functions for generating decision trees
  • decision_tree_traverse.py: functions for traversing a decision tree to predict labels
  • density_tree.py: data struture for density tree nodes
  • density_tree_create.py: functions for generating density trees
  • density_tree_traverse.py: functions for descending density trees and retreiving their Gaussian parameters
  • density_forest.py: functions for creating density forests
  • helper.py: helper functions
  • plots.py: functions for plotting the data
  • random_forests: functions for creating random forests

Supervisors:

  • Prof. Devis Tuia, University of Wageningen
  • Diego Marcos González, University of Wageningen
  • Prof. François Golay, EPFL

Cyril Wendl, 2017