search

hongzimao / input_driven_rl_example

Variance Reduction for Reinforcement Learning in Input-Driven Environments (ICLR '19)
https://people.csail.mit.edu/hongzi/var-website/index.html
MIT License
31 stars 10 forks source link

some questions about meta-learing approch #2

Open Suliucuc opened 4 years ago

Suliucuc commented 4 years ago

Hello, I have read your paper carefully.The paper mentioned that in addition to the method of multi-value network, there is also the method of meta-learning. Could you please share the code of meta learning?I would appreciate it very much.This is very urgent for me,So I really hope you can reply to me as you see this issue.Thank you again.

hongzimao commented 4 years ago

Hi, I'm attaching the meta-learning code here. Please note that I pulled the code from the middle of a (long-delayed) refactoring process -- the code might be messy and you might want to play with the parameters (especially lr_rate and meta_lr_rate in the MAML constructor in load_balance_actor_maml_critic_train.py).

After saving the attached two files (in .zip) in the input_driven_rl_example directory. Run

python3 load_bance_actor_maml_critic_train.py --num_workers 10 --service_rates 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95 1.05 --result_folder ./results/maml_results/ --model_folder ./results/parameters/maml_results/

I'm also attaching the testing learning curve from my run of this code.

Hope these help!

maml_code.zip

maml_test_performance