Notes on Optimal Stopping

 

Status: Just getting started studying this.

 

Summary:

 

Conjecture: The most common applications of optimal stopping algos are similar to the Secretary Problem.

 

That problem basically asks "When do we stop evaluating interviewees?"  and is equivalent to a whole lot of more interesting things, such as "When do we stop evaluating employment offers?"

 

Some solutions are known that offer at least a 33% chance of getting the right answer, such as:

 

1. You have N applicants, and e is the log number e
2. Reject the first N/e candidates  
3. After that, stop at the first applicant that is better than every applicant so far.

 

Of course, that is an oversimplification and leaves a lot of considerations out.

 

Wikipedia Clickhole References:

 

Read through these stopping when you can no longer grok the math (optimal! ;) and you will have a good overview:

 

Optimal Stopping

Stopping Rule

Secretary Problem

Odds Algorithm