- 11 Jan, 2020 4 commits
-
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
Since we disable XML, we also need to stop storing the games.
-
- 10 Jan, 2020 2 commits
-
-
Vivien Kraus authored
Resolve "Review the code generation for the card names" Closes #179 See merge request !183
-
Vivien Kraus authored
The function was huge. This was not a good idea because frama-c could not say anything. I do not know if it will be better for frama-c, but I can tell you it is more readable that way.
-
- 07 Jan, 2020 1 commit
-
-
Vivien Kraus authored
This is not a real dependency.
-
- 06 Jan, 2020 3 commits
-
-
Vivien Kraus authored
Develop See merge request !182
-
Vivien Kraus authored
The fuzziness of the AI is expressed as summed up for all players, because each random bid or slam declaration results in an unbalanced game which is uninteresting.
-
Vivien Kraus authored
-
- 03 Jan, 2020 4 commits
-
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
- 02 Jan, 2020 10 commits
-
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
It does not matter because =tarot_game_free= calls =free= anyway, but the type was not correct.
-
Vivien Kraus authored
We learn from a batch of 100 games at a time.
-
Vivien Kraus authored
Now that we changed year, it is obvious that it does not work.
-
Vivien Kraus authored
The goal is to be able to bootstrap the package from a docker image containing these specifications.
-
Vivien Kraus authored
Since the commit is fixed, it is not the latest commit so it is not present if we only shallow clone it.
-
Vivien Kraus authored
-
Vivien Kraus authored
-
- 29 Dec, 2019 3 commits
-
-
Vivien Kraus authored
-
Vivien Kraus authored
It is very very rarely used, and not entertaining.
-
Vivien Kraus authored
Not having a call is uninteresting.
-
- 19 Dec, 2019 13 commits
-
-
Vivien Kraus authored
With a barycenter between pure and random, when there are only 2 options (pass vs push, for instance), there is never any random.
-
Vivien Kraus authored
player The AI is not trained to take, or keep.
-
Vivien Kraus authored
-
Vivien Kraus authored
Since there are no numbers on the corners, we can't read them when in our hand.
-
Vivien Kraus authored
I noticed because the pixmaps were not correctly drawn.
-
Vivien Kraus authored
Or at least, an early version that is in the public domain.
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
-
Vivien Kraus authored
In the early stages of the game, a pure perceptron will have a very slim amplitude for the predictions, because anything can happen. However, the noise will still be between -100 and 100. By this trick, all predictions have a mean of 0 and a standard deviation of 1.
-
Vivien Kraus authored
If all cards yield the same prediction, then at least we will avoid boringly playing the hearts, then the clubs, then the diamonds, then the spades.
-