CODEWITHSUNDEEP
algorithmmachine-learningartificial-intelligence
user2951186
user2951186

Reputation: 19

What is the name of this learning algorithm?

I designed a learning algorithm for an AI player that tries different methods of winning the game and then continues on with method that has the highest win rate. If the highest win rate method starts to lose it goes onto the method with the 2nd highest win rate etc...

I need to do a write up on the learning algorithm, what could it be called?

Upvotes: 0

Views: 205

Answers (3)

Raff.Edward
Raff.Edward

Reputation: 6544

The algorithm has no name because its not particularly good (no offense) relative to what exists for such problems.

What you are "solving" is the Multi-armed Bandit problem. The "different methods of winning" are your slot machines, and your "win rate" is the payout odds of each slot machine. Because you are playing against another person and they might change their strategy, you must deal with the possibility that the best slot machine might "start to lose". This is an adversarial scenario. There are many papers and algorithms on this topic from many different perspectives / scenarios. Its very well studied and people are still interested in it.

I assume this is for some class project report. I would simply call it "My heuristic approach", because that is what it is.

Upvotes: 5

ChrisProsser
ChrisProsser

Reputation: 13118

That probably depends on how you have implemented it. It sounds a little like a multi-class logistic regression algorithm. This evaluates the probability of a given outcome for each of the classes and then chooses the best probability, but this could also be implemented with neural networks or various other algorithms.

Upvotes: 1

jean
jean

Reputation: 4350

It's a classical Try & Error approach. TryError, TryAndError, TaE, Try&Error-o-matic, pick whatever you like more.

Upvotes: 1

Related Questions