Summer work on mean-field inference at EPFL CVLab
  • Jupyter Notebook 82.5%
  • Python 17.5%
Find a file
2017-08-16 09:00:03 +02:00
.gitignore changes 2017-07-04 11:52:40 +02:00
create.py --single option for unique solution grids creation 2017-08-10 16:14:10 +02:00
discr.py great stuff 2017-08-01 16:37:56 +02:00
experience.py generate learning curves 2017-08-14 18:31:43 +02:00
judge.py argument parse only if not used as a library 2017-08-16 08:59:25 +02:00
learning.py avancées 2017-07-06 10:47:28 +02:00
Meanfield learning.ipynb great stuff 2017-08-01 16:37:56 +02:00
mf.py a log2 -> log fix + optimized annealing 2017-08-14 11:54:02 +02:00
README.md Added some description 2017-08-02 19:52:09 +02:00
sudoku.py some comments 2017-08-10 16:17:58 +02:00
sudoku_learn.py pre-learning feature 2017-08-14 11:54:24 +02:00
sudoku_learn_cnn.py add option to judge on another dataset 2017-08-16 09:00:03 +02:00
sudoku_test.py avancées 2017-07-06 10:47:28 +02:00
Test mean field.ipynb Multi-filter mean-field sample 2017-08-03 16:15:17 +02:00
test.py modified this 2017-07-14 15:00:46 +02:00
Untitled.ipynb great stuff 2017-08-01 16:37:56 +02:00

On Mean-Field inference and machine learning: the example of Sudoku.

Work done during the Summer 2017 internship at the CVLab - EPFL.

  • create.py: create a sudoku dataset, you can specify the size of the board, the number of examples and the difficulty of the dataset.
  • discr.py: this script tests the discriminative power of a given CRF (not updated to version 2.0)
  • experience.py: launches a bunch of experiments (not updated)
  • judge.py: evaluates the quality of a learned CRF against a dataset, specifying the number of modes (in case of multi-modal mean-field).
  • learning.py: some MF samples (not updated)
  • mf.py: the core MF/MMMF inference library
  • sudoku_learn.py: uses the MMMF inference to learn the game of sudoku using a multi-modal gradient method.
  • sudoku.py: some utility library in order to handle sudoku dataset.