Copy the files 
  Number_Generators.f90
  Minerva.f95
  Forgetting.f95

to your directory. Minerva.f95 will simulate retrieval of a probe given a label, or retrieval of a label given a probe (classification). 

To compile, type: 

f95 Number_Generators.f90 Minerva.f95 -o Minerva

and to run, just type "Minerva" at the prompt. You can play around with the parameters and output, and switch the category that is classified. 

Forgetting.f95 builds the same memory based on Minerva, but it manipulates the delay time and observes the effect on retrieval. Try manipulating the F parameter (probability that a feature reverts to zero) over a smaller time scale, say, from 0 to 0.9 in increments of 0.01. You'll see that as the time delay increases, classification of the prototype (never stored) becomes actually better than an old exemplar (actually stored). Although this is not a property we would want in a computer memory, it is an adaptive function of human memory over time. 

To compile, type: 

f95 Number_Generators.f90 Forgetting.f95 -o Forget

and to run, just type "Forget" at the prompt. Play around wit the forgetting (delay) parameter to examine its effect on classification performance or on retrieving a pattern given a label over time.