OCR SUDOCUL

A program that solves sudokus with a photo of the grid.

BUILD THE PROJECT

git clone https://github.com/Michael-Rousseau/OCR_sudocul.git cd OCR_sudocul Use the command make at the root of the project.

CLEAN THE PROJECT

make clean at the root of the project.

EXECUTABLES

All executables can be found in the build folder (at the root).

DOCMENTATION

XOR

NAME

• xor - play with a XOR neural network
  

SYNOPSIS
- ./xor [network options...] mode [mode options...] [-s] GENERAL CASE

- network options:
- - -f path (IMPORT NETWORK FROM FILE)
- - EXAMPLE: ./xor -f network.nw mode [mode options...] [-s]

- - -lh [layer size ...] (CREATE NETWORK WITH CUSTOM DIMENSIONS)
- - EXAMPLE: ./xor -lh 12 8 mode [mode options...] [-s] (2 HIDDEN LAYERS)

- mode (ONE OF THE FOLLOWING):
- - test nb_tests
- - - - nb_tests: integer (number of tests to make)

- - learn nb_tests learning_rate
- - - nb_tests: integer (number of tests to make)
- - - learning_rate: double (learning rate of the network)

- xor [network options...] mode [mode options...] -s [path]
- - SAVE THE NETWORK AFTER EXECUTION (TO path IF SPECIFIED)

DESCRIPTION
- xor is an easy to use XOR neural network. You can dimension it as you like,
- easily perform tests and make your network learn. You can also import and
- export your network easily.

EXAMPLES
- ./xor -lh 12 learn 10000 0.9 -s
- - Create a XOR neural network with a single 12 neuron hidden layer, make
- - it learn with 10000 tests with a learning rate of 0.9, and save it to
- - the default file: 'network.nw'
- ./xor -f network.nw test 100000
- - Load a XOR neural network from the network.nw file and perform 100000
- - tests on it.