1 of 6

RL4Sys Framework Project

Jeffrey Wang

Dr. Dong Dai

2 of 6

Motivations

  • Missing: a plug-and-play deep reinforcement learning framework
  • Example applications: smarter job scheduling, memory allocation
    • Benefit of this framework
  • Results: a useful tool for public use, personal knowledge

3 of 6

Backgrounds

Literature

  • David Silver ML Course
  • Winder - Reinforcement Learning

Tools

  • Spinning Up

Concepts

  • Reinforcement learning terminology
  • Types of RL, model and model-free
  • Policy Gradients

​

4 of 6

Current Progress

  • Gained background knowledge
  • Experimented with Spinning Up agents
  • Implementations of reference learning
    • Dynamic Programming
    • Tabular Q-learning
  • Beginning to work with research components

5 of 6

Future Plans

  • Learn to write tabular RL algorithms: Spring break (Mar 1)
  • Make contributions enabling RL4Sys for specific use cases (March)
    • Job scheduling (Mar 1-15)
    • Data structure optimization (Mar 15-30)

6 of 6

Thank You

Q & A