Go to any corner to adjust learning speeds and number of boxes. The learning speeds are adjusted while running the project, but the number of boxes are reset only after restarting the project or when the project resets if it detects too much lag. Be warned though: Learning fast doesn't always lead to a better outcome! Also if you want to enable fast mode, it is in the bottom right corner.
(NOTE: You can find how to interpret this project under 'game play'.) Wow this is getting a bit over-rated. I have always thought of hide and seek as a childish game played by very young children. However, as I have analyzed it from a mathematical point of view, it is a more complex game than I thought. For starters, it is a game with no best strategy, meaning that every strategy is only useful if no one else uses it. A secret hiding place, for instance. Also, it requires the person finding people to guess where they are most likely to be. That is far more complicated than most of us realize. So I decided to simulate two bots taken on the challenge to sit for eternity playing 'Box Hide and Seek' GAME PLAY: In every round, the hider hides an object inside a box. The hider's brain shows how likely it is to pick a certain box. The darker the box, the more likely it is to pick that one. Once the hider picks a box, it shows what it picked on the row labeled 'box'. Then, the finder guesses the location of the object, shown in the 'guess' row. It's brain also shows how likely it is to pick certain boxes. After each round, the outcome is calculated. If the finder picks the right box, then the finder wins and the hider loses. Vice versa if the finder doesn't pick the right box. Both the finder and hider learn from their past rounds. If they lost, then they would reduce the chance of making the choice that made them fail by a bit, and increase the chance of picking other boxes. If they succeeded, they would do the opposite: Making it more likely to pick their past choice, and less likely to pick other choices. A bit of randomness is also added so that they do not get stuck for way too long in specific brain configurations (ok, that sounds weird.) unless it was the best strategy... so far. FEATURING: Probability distributions. Normalization. Displaying text-like stuff with pen (the grid squares are placed one after the other, like in text.)
