More Training Data for Exam Preparation [Alpha-Beta-Pruning]

Disclaimer: Dieser Thread wurde aus dem alten Forum importiert. Daher werden eventuell nicht alle Formatierungen richtig angezeigt. Der ursprüngliche Thread beginnt im zweiten Post dieses Threads.

More Training Data for Exam Preparation [Alpha-Beta-Pruning]
Hi,

I am not very creative so I generated some trees to practice alpha beta pruning under time pressure. Maybe someone finds this helpful.

Pruned nodes are red in the solution.

Without solutions: https://www.tomabrafix.de/ki/alphabetapruning-print.pdf
With solutions: https://www.tomabrafix.de/ki/alphabetapruning.pdf

Large random shit: https://www.tomabrafix.de/ki/alphabetapruning_largeset.pdf
(I did not really look at these)

All tree files (dot and pdf): https://www.tomabrafix.de/ki/trees

If you find a bug (attach seed pls) or want other tree structures just answer here.

4 „Gefällt mir“

I think the solution for 1927540640805326515 (the first tree of random shit) is wrong, the algorithm should prune I, J, L and M, not just L and M. Maybe somewhere you have a “less than” where there should be a “less or equal.”


No, I think it is correct. C is the Max player. The Max player tries to maximize alpha and prunes if he finds a value which is greater or equal than beta. So in this tree when alpha-beta-pruning visits C it has alpha=-inf and beta=0. And therefor nodes get only pruned if there is a child which is greater or equal to 0. You can also imagine that if J had a value like +42 it would be inacceptable for the Min player at node A to select C because the Max player then has the opportunity to achieve +42 instead of 0 at B.


Sorry, disregard what I said. I didn’t notice the game started with Min.


Here is the solution for the same tree with Max player starting: https://www.tomabrafix.de/ki/trees/tree_1936547840060067507_sol.dot.pdf