Bank vault holdup webapp

“Bank vault holdup” is a game where you have to guess a secret pass to open the bank vault and get the reward! There is no hint, you just have to follow your intuition.

Click on the screen to launch the game in your browser. The loading mays take some seconds.

screenLet me know if some bugs appear in your browser (I know there is a font problem with IE, and the game does not work on cellphones).

Continue reading

Let a software control your keyboard and your mouse

outputSikuli is a software that can control your mouse and your keyboard, and allows to do some automatic tasks.
It is useful when:

  • those tasks cannot be done through an API (e.g. for flash websites),
  • when you don’t want to learn an API (e.g.for a single work).

Sikuli works on all platforms and is available here.

In the sequel, I show an example to capture images from Google Maps and make a map from them. The aim is to make a draw of a French island, l’île d’Oléron.

Continue reading

Introduction to particle filters: a trajectory tracking example

Those slides are an introduction to particle filters, with an homemade example of trajectory tracking. The presentation has been given in 2014 for the Colloque Inter’Actions 2014.

I think it’s preferable to have read the previous presentation (on discrete hidden Markov models) before this one (on continuous hidden Markov models).

In the following, I describe two important algorithms for particle filtering: the Sequential Importance Sampling algorithm (SIS) and the Sequential Importance Sampling Resampling algorithm (SISR) . The example is here to understand how work the algorithms and to focus on the difference between them, especially through the weights degeneracy problem.

Continue reading

Introduction to hidden Markov models

I show here my slides of a talk I gave for doctoral students in 2013. It is an introduction to hidden Markov models (on finite state spaces), following the tutorial of L. R. Rabiner.

I explain what is a hidden Markov model and what are classical issues with those objects.Then, some algorithms are taken into consideration: forward decomposition, Viterbi algorithm and a simple EM algorithm.

Those algorithms are illustrated step by step through the following homemade toy example. A consumer walks everyday between three places A, B and a shopping center C. A marketer standing at C, only gets a noisy information about the position of the consumer. He tries to identify where the consumer is located.

Continue reading

Gender of French nouns

genderLetterIntroLike other Romance languages, French nouns are divided in two almost distinct genders: feminine and masculine. Deciding the gender of a word is not obvious. For example, a table is feminine (“une table”) but an acid is masculine (“un acide”).

On the following plots, we check out the distribution of the gender across the letters. We then learn a simple tip to choose the gender of an unknown noun.

Continue reading