Great explanation! Some important notes as well: 1) only in a Max node can update the corresponding alpha, so does Min for beta. 2) v can only be returned up to its parent 3) alpha and beta can only be passed down from its parent 4) cut the current node from the tree whenever alpha >= beta In the graph on the right whiteboard, shouldn't all the alpha, beta are initialized as -infinity and infinity, respectively? At 13:07, for the completeness of the logic flow, I think we should set the beta to be 2 in the node(v=2, alpha=8, beta= ), then because alpha(8) > beta(2), we prune the right subtree.
Every single time I see that you have uploaded a video about a topic, I lose all fear, that I'm not going to be able to understand it. This is not an exaggeration, you've helped me avoid so much confusion and desperation. Thank you so much John!
Great teacher, got a B on my AI exam with your videos about tree search algorithms 👌 explained it way better in a couple of minutes than my teacher in an hour
Wonderful video. This helped a lot! Working through the example in real time and explaining your thought process as you went really helped make this click.
the best explanation about minimax. You are a great professor. I watched a lot of videos about minimax algorithm, but only your video I understood clearly
I want to thank you sir, you have done what my teacher was unable to do, a good explanation, it took me three days to understand this, and finally I understood thanks to this video
Watched it in approximately 9 minutes (X1.5 speed) and this has been the most precise and to-the-point explanation I have watched so far and believe me I have watched many. Thank you sir
Thank you sir! I was wondering if there is a formula that can calculate the maximum number of nodes that can be pruned in a tree with a depth of d and a branching factor of b.
Wow, this lecture helps me understand the alpha-beta algorithm, the pseudo code is very helpful to understand. Quite intuitive. Where is the link for the assignment?
The best explanation of alpha beta pruning. "If the value is greater than beta that means min is going to prefer the branch that lead to beta." That line made it click for me.
Is it at all possible to see the slides you mention in the video? I've spent the past couple of days trying to figure out how to apply MiniMax with Alpha-Beta Pruning to my A-Level Chess AI, and I always seem to get bugs that shouldn't exist (I literally mean they shouldn't exist, not I don't know why they exist), Thanks
I must be missing something because the following scenario seems like a possible scenario where alpha beta pruning can discount a superior position down the road. In chess, isn't it possible to have a move sequence where moves 6 ply out may show an inferior score for white but if the line was analyzed further, we'd see that white might be able to force a mate by taking a material loss. Wouldn't alpha beta pruning possibly discount such a line by abandoning further computational depth for that line such that it would never be able to see that a temporary drop in maximization of white's score actually sets up for a win for white? I guess my question can be reduced to this: Is it possible for alpha beta pruning to discount a line that is temporarily inferior (low score) but a few moves further out we discover a forced mate? Could alpha beta pruning throw out the analysis for such lines? Also, the entire principle of the minimax algo (in my opinion) is the scoring we give specific positions based on what information has been learned about chess over the centuries. We rank certain material with specific scores and also certain positions with specific scores. If the methodology of scoring the current state of the chess game is imperfect, then the minimax algo may not always be 100% optimal in the search tree it explores based on those scores. Let me give a quick counter example to (hopefully) better illustrate my argument. In the game of chess, there is really only one ultimate score -- is there a sequence of moves from the current game position where white (or black if playing black) can force a checkmate of the opposing king. Every other analysis within the game is of no consequence if it doesn't lead to a line where the opposing king is checkmated (or if checkmating is not possible, the only other analysis of consequence is trying to draw the current game). A new chess engine called Oracle is created using hardware from thousands of years into the future. Oracle is able to analyze 100 ply within the game in a matter of microseconds. Oracle discovers a line where white sacrifices four pieces (let's say a pawn, a knight, a bishop and a rook) in order to force a mate on the opposing king approximately 20 moves out. Now in this example, wouldn't current engines stop processing this line after the first or second piece was lost because the minimax score / alpha beta pruning? The current engine might discount the sequence of moves and call the first sacrifice a blunder because it never processed this particular line deeper. I could be way off here because my knowledge of minimax with A/B pruning is incomplete -- that's why I'm throwing this out there for others with more experience. Is this a disadvantage of minimax with AB pruning?
Great explanation! Some important notes as well:
1) only in a Max node can update the corresponding alpha, so does Min for beta.
2) v can only be returned up to its parent
3) alpha and beta can only be passed down from its parent
4) cut the current node from the tree whenever alpha >= beta
In the graph on the right whiteboard, shouldn't all the alpha, beta are initialized as -infinity and infinity, respectively?
At 13:07, for the completeness of the logic flow, I think we should set the beta to be 2 in the node(v=2, alpha=8, beta= ), then because alpha(8) > beta(2), we prune the right subtree.
all solid things to point out, thanks
It works like a charm, Thanks!
Yeah, this algorithm is total shit
beta would never assigned the value of two; the function would return before that variable assignment.
Thanks this helped me so much, you have no idea
Excellent explanation! The video is clear, concise, and precise. We want more videos from you. Please teach us more about AI.
You are a great lecturer, not many people can do this.
Thanks! :)
Just wanted to thank you for your great videos! Watching them pre-exam to refresh and they're all very clear and informative
This definitely has to be the most thorough explanation I found on the topic. Thank you! :D
Every single time I see that you have uploaded a video about a topic, I lose all fear, that I'm not going to be able to understand it. This is not an exaggeration, you've helped me avoid so much confusion and desperation. Thank you so much John!
THANK YOU SIR! I finally know alpha beta pruning!!! Your explanation was crystal clear. If you taught my AI course, I would never ditch class!!
As a self-learner, this video is by far the best explanation I've ever seen. Clean, concise, and to the point.
it's just amazing. all your videos are the definition of crystal clear explanation. perfect. I wish you'll do more videos
Great teacher, got a B on my AI exam with your videos about tree search algorithms 👌 explained it way better in a couple of minutes than my teacher in an hour
Great work! Well done
Wonderful video. This helped a lot! Working through the example in real time and explaining your thought process as you went really helped make this click.
Great explanation! I already knew what steps to take, but it wasn't really clear why I should take those steps. It is clear to me now.
Respect to you sir. You explained it so well! I'm very grateful!
One of the best video on youtube to understand Alpha Beta pruning.
Want more video like this to get more understanding of AI.
Your videos are so concise and helpful. They're the only thing keeping my grade decent in my Intelligent Systems class.
Glad you like them!
the best explanation about minimax. You are a great professor. I watched a lot of videos about minimax algorithm, but only your video I understood clearly
5 year later and still best video on alpha-beta pruning. Thank you sir.
Thanks for making me feel old! Very glad it's useful 😀
This is the best video I have seen which explains the algorithm along with the graph.
He is god. He explained everything in just 13 minutes. I wish if I came across this video earlier it could have saved my time. Great teacher!!
Your videos help me so much with my course, thanks!
Awesome!! Seen most of your videos(Loved A* video too) and I must say that this topic was a bit tricky till I saw this ! You make it sound so simple
Watching this once saves me so much time that might've gone into reading through a textbook section on this particular algorithm
Thanks. That was my nth video on that topic and this time it clicked. Thanks for taking the time to go through it so deliberately. :)
You're very welcome!
Can you do a video on nonlinear planning with Goal Stack...
One of the best explanation, Thank you John Levine.
I want to thank you sir,
you have done what my teacher was unable to do, a good explanation,
it took me three days to understand this, and finally I understood thanks to this video
Nice lesson, Biblo. The great eye watching you is pleased.
Thank you Sir, after checking many videos on the same topic, your lecture makes me understand the topic.
He saved my exam. Thnxx sir for such a great explanation.
Thank You so much sir!
God Bless you!
Thank you. I've seen many videos and could not understand until I saw this video. Many thanks for your clear explanation!!
this is amazingly helpful for me! thank you sir for this explanation!
Fantastic explanation. No one on CZcams has ever explained this algorithm so nicely using it's pseudo code . Very nice Sir, Respect
Watched it in approximately 9 minutes (X1.5 speed) and this has been the most precise and to-the-point explanation I have watched so far and believe me I have watched many. Thank you sir
Brilliant explanation. Just what I needed. Thank you.
This is the best video about this topic
thank you for the video, it really helped me a lot doing my exam study!!
Don't expect to instantly understand this topic, but your video did it. Thanks. :D
Really good!! Thanks for taking the time to explain it properly.
Thank you, sir. You are a great teacher!
Thank you for this amazing video and explaination, not many have the ability to do this 🌸🖤
Thank you sir! I was wondering if there is a formula that can calculate the maximum number of nodes that can be pruned in a tree with a depth of d and a branching factor of b.
Thank you very very much! It takes a long time for me to understand the alpha-beta pruning .Luckly ,I find this vedio.
Would the -2 value not be pruned as we have a deep cut off situation ? We can see that Max will be = 8? Is there point in exploring -2?
an excellent teacher , thank's a lot. we'r waiting for more courses about AI :) Good luck
This is so helpful!! Thank you so much!
Thank You . You're an amazing teacher .
Thanks Shivam! Glad you found it useful.
Thank you Dr
Thank you so much! You're the best!!
Great presentation of work. would it be possible to link the supplementary material associated with this lecture?
Damn, you are good. Thanks for helping
Thank you so much!
Wow, this lecture helps me understand the alpha-beta algorithm, the pseudo code is very helpful to understand. Quite intuitive. Where is the link for the assignment?
I have to admit you are the only one who could explain and impart knowledge to even the dumbest person like me. You are the great great teacher.
Amazing videos! Thank you!
Goated Explanation Sir!
Best video I've seen on the subject
thank you so much for the lecture.
The only explanation that made sense to me!
Thank you so much for your video. It's really clear.
Great video, great explanatiom!
Great explanation!
Awesome explanation sir thanks from india
Thank you!!! :)
loved the way you explaine
Good work right there!!!
The best explanation of alpha beta pruning. "If the value is greater than beta that means min is going to prefer the branch that lead to beta." That line made it click for me.
great explanation!!! Thank you!!!
this video was so helpful.
thank you.
very good Explanation Sir! :)
This is amazing. Thanks!
this is life-saving sir thanks a lot ^-^
OMG! Very very clear! I came from CS188 step by step.
Thank you so much sir. !!!!
Please make a video on Negamax as well, would be perfect if includes Transposition in the algorithm either. And thanks for the lesson, thank you.
Is it at all possible to see the slides you mention in the video? I've spent the past couple of days trying to figure out how to apply MiniMax with Alpha-Beta Pruning to my A-Level Chess AI, and I always seem to get bugs that shouldn't exist (I literally mean they shouldn't exist, not I don't know why they exist), Thanks
Very clear explanation, Thank you
Great explanation 👌
Awesome video!
bunch of thanks (y)
Thank you!
Thanks a lot, you made my day !
I must be missing something because the following scenario seems like a possible scenario where alpha beta pruning can discount a superior position down the road.
In chess, isn't it possible to have a move sequence where moves 6 ply out may show an inferior score for white but if the line was analyzed further, we'd see that white might be able to force a mate by taking a material loss. Wouldn't alpha beta pruning possibly discount such a line by abandoning further computational depth for that line such that it would never be able to see that a temporary drop in maximization of white's score actually sets up for a win for white?
I guess my question can be reduced to this:
Is it possible for alpha beta pruning to discount a line that is temporarily inferior (low score) but a few moves further out we discover a forced mate? Could alpha beta pruning throw out the analysis for such lines?
Also, the entire principle of the minimax algo (in my opinion) is the scoring we give specific positions based on what information has been learned about chess over the centuries. We rank certain material with specific scores and also certain positions with specific scores. If the methodology of scoring the current state of the chess game is imperfect, then the minimax algo may not always be 100% optimal in the search tree it explores based on those scores.
Let me give a quick counter example to (hopefully) better illustrate my argument. In the game of chess, there is really only one ultimate score -- is there a sequence of moves from the current game position where white (or black if playing black) can force a checkmate of the opposing king. Every other analysis within the game is of no consequence if it doesn't lead to a line where the opposing king is checkmated (or if checkmating is not possible, the only other analysis of consequence is trying to draw the current game). A new chess engine called Oracle is created using hardware from thousands of years into the future. Oracle is able to analyze 100 ply within the game in a matter of microseconds. Oracle discovers a line where white sacrifices four pieces (let's say a pawn, a knight, a bishop and a rook) in order to force a mate on the opposing king approximately 20 moves out.
Now in this example, wouldn't current engines stop processing this line after the first or second piece was lost because the minimax score / alpha beta pruning? The current engine might discount the sequence of moves and call the first sacrifice a blunder because it never processed this particular line deeper.
I could be way off here because my knowledge of minimax with A/B pruning is incomplete -- that's why I'm throwing this out there for others with more experience. Is this a disadvantage of minimax with AB pruning?
Great Explanation! Thanks alot.
Very good video. I will now attempt to code my own tic-tac-toe solver :)
Sirrr love your vids
Many thanks for explanation!
great explaining prof
Do we always start with the root node as the maximum one?
thanks man, you gave me a great favor
Thanks!
Thanks teacher...I watch your videos in Azerbaijan✨
Thank you sir ！
Cheers mate thanks!!
great explanation!! thanks a lot
it is a really nice explanation. thank u sir.
this mans single handedly getting me through AI
Awesome explanation. Upvote. Please make videos explain, all alogrithms Thanks!
Sir, would you mind making a series video for Artificial Intelligence... It will be more helpful...
Awesome Sir, @respect
Awesome, very clear