Archive for Liberation

non, toujours non

Posted in Kids, pictures with tags , , , , on May 3, 2017 by xi'an

ne rien lâcher

Posted in Statistics with tags , , , , on April 25, 2017 by xi'an

threats on abortion rights

Posted in Kids, pictures with tags , , , , , , on December 3, 2016 by xi'an

231116_yesuAfter a surrealistic debate last week between the two primary candidates about their comparative proximity to the catholic pope, and the reminder by the winner, François Fillon, that he did not consider abortion as a fundamental right, the French catholic church hierarchy has been trying to block a new law that criminalises anti-abortion websites that mimic as official websites, while providing misleading information… With the usual newspeak deviance of arguing for a freedom of speech that is not under threat.

Libération [hardcover]

Posted in Books, pictures with tags , , , on November 23, 2016 by xi'an

tatchois

Libération [cover]

Posted in Books, pictures with tags , , , on November 7, 2016 by xi'an

Libération plays with Lego

Posted in Kids, pictures with tags , , , on April 10, 2016 by xi'an

sudoku break

Posted in pictures, R, Statistics with tags , , , , on December 13, 2013 by xi'an

sudo291113While in Warwick last week, one evening after having exhausted my laptop battery, I tried the following Sudoku (from Libération):

>   printSudoku(readSudoku("libe.dk"))
  +-------+-------+-------+
  | 4   6 |   2   | 3   9 |
  |   3   |       |   2   |
  | 7   2 |       | 5   6 |
  +-------+-------+-------+
  |       | 9 4 5 |       |
  | 5     | 7 6 2 |     1 |
  |       | 3 1 8 |       |
  +-------+-------+-------+
  | 6   9 |       | 1   3 |
  |   7   |       |   9   |
  | 3   1 |   9   | 4   7 |
  +-------+-------+-------+

and could not even start. As it happened, this was a setting with no deterministic move, i.e. all free/empty entries had multiple possible values. So after trying for a while and following trees to no obvious contradiction (!) I decided to give up and on the next day (with power) to call my “old” sudoku solver (built while at SAMSI), using simulated annealing and got the result after a few thousand iterations. The detail of the exploration is represented above, the two colours being code for two different moves on the Sudoku table. Leading to the solution

  +-------+-------+-------+
  | 4 8 6 | 5 2 1 | 3 7 9 |
  | 1 3 5 | 6 7 9 | 8 2 4 |
  | 7 9 2 | 8 3 4 | 5 1 6 |
  +-------+-------+-------+
  | 2 1 3 | 9 4 5 | 7 6 8 |
  | 5 4 8 | 7 6 2 | 9 3 1 |
  | 9 6 7 | 3 1 8 | 2 4 5 |
  +-------+-------+-------+
  | 6 2 9 | 4 8 7 | 1 5 3 |
  | 8 7 4 | 1 5 3 | 6 9 2 |
  | 3 5 1 | 2 9 6 | 4 8 7 |
  +-------+-------+-------+

I then tried a variant with more proposals (hence more colours) at each iteration, which ended up being stuck at a penalty of 4 (instead of 0) in the final thousand iterations. Although this is a one occurrence experiment, I find it interesting that having move proposals may get the algorithm stuck faster in a local minimum. Nothing very deep there, of course..!

sudo301113