10‐31‐2023 Weekly Tag Up

Attendees

• Chi-Hui
• Joe

Updates

• New experiments to run:
  • Add number of rounds from 7 to 20
  • Run 4-agents experiments (no copy)
  • Run 9-agent experiments (no copy)
  • Run 16-agent experiments (no copy)
  • NOTE: in order to run the 9 and 16 agent experiments, new baseline policies must be generated (the queue and threshold policies)
• Chi-Hui to give Joe access to lab's desktop
• AAMAS response should come near start of December
• Experiment enhancements
  • Use learned policy to enhance dataset
  • Goal is to get single learned policy to approach the performance of the individual "queue" and "threshold" policy
  • Perform dataset updates every couple rounds
• Enhancements are lower priority than other experiments
  • Can explore these more after we hear back AAMAS

Logistics

• Keep meeting at 5pm MST for now (may need to change later)