Randomizing Rewards

 

The screen changes colors at times forming a Poisson process with a rate specified by the user. The times between these events are independent, exponentially distributed random variables; hence, they have the "memoryless property," which means that the times cannot be anticipated. This should be useful in animal training, e.g. sit/stays.

Enter the mean time between epochs, and click on "Start" to start. Whenever the screen color changes, reward if your animal's behavior is desirable. If you cue a stay and your dog doesn't, simply reset the stay and reward at the next screen color change; you do not need to restart the program or wait for a screen change to reset your animal. (Due to the memoryless property of the exponential distribution, the time until the next screen change has the same distribution no matter how long since the last screen change.) When done, hit stop.

Note that around 63% of the times will be less than the mean.

 

Enter mean time between changes in seconds:
(180 seconds = 3 min., 300 seconds = 5 min., 3600 seconds = 1 hr.)

Press start when starting the experiment:

Press stop when finished:

Change desired success rate? (e.g. .8 or .85):

Enter Number of successes:

Enter Number of failures:

Suggested new mean:

Update mean (otherwise the old mean is used):

Either hit continue to start a new experiment:

or hit done if finished:

 

 

Poisson randomizer, created by Robert D. Foley
e-mail: rfoley@isye.gatech.edu