mo8it/ReCo.jl
1
0
Fork 0
mirror of https://gitlab.rlp.net/mobitar/ReCo.jl.git synced 2025-09-02 09:02:35 +00:00
Code Issues Releases Activity

Fixed normalization

Browse source
This commit is contained in:
Mo8it 2022-01-30 03:32:47 +01:00
parent 02739b7de6
commit eb29ca3d10
3 changed files with 10 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

9
analysis/reward_discount_analysis.jl
View file

@ -14,13 +14,14 @@ function run_reward_discount_analysis()
γ = γs[γ_ind]
env_helper = ReCo.run_rl(;
EnvType=ReCo.OriginEnv,
n_episodes=500,
episode_duration=8.0,
n_particles=200,
n_episodes=3,
episode_duration=15.0,
n_particles=150,
update_actions_at=0.08,
ϵ_stable=0.00001,
process_dir="reward_discount_analysis/$γ_ind",
reward_discount=γ,
show_simulation_progress=false,
)
rewards = env_helper.shared.hook.rewards
@ -34,7 +35,7 @@ function run_reward_discount_analysis()
ax = Axis(fig[1, 1]; xlabel="Episode", ylabel="Reward")
rewards_plots = []
for (rewards_ind, rewards) in enumerate(γ_rewards)
for rewards in γ_rewards
rewards_plot = lines!(ax, 1:length(rewards), rewards)
push!(rewards_plots, rewards_plot)
end