Phases of Simulated User Experiments

  • Evolution
    • The classification performance should be improved as the number of examples per category increases while NO new categories are introduced.
  • Recovery
    • By increasing the number of categories, it is expected that the prediction accuracy decreases. The time spent in system evolution until correcting and adjusting all current categories defines recovery.
  • Breakpoint
    • Eventually the learning agent reaches to a breakpoint where the agent is no longer able to learn more categories.