blakeelias / pandemic_RL

Reinforcement learning for economically optimal pandemic response.
GNU General Public License v3.0
2 stars 1 forks source link

Discrepancy in policy plots #49

Open blakeelias opened 3 years ago

blakeelias commented 3 years ago

New: http://ec2-3-138-203-155.us-east-2.compute.amazonaws.com:8001/results/env%3D%28R_0%3D2.5%2Cdistr_family%3Dpoisson%2Cimported_cases_per_step%3D0.0%2Cdynamics%3DSIS%2Ccustom%3D%7B%27extra_scale%27%3A%201.0%2C%20%27tags%27%3A%20None%2C%20%27time_lumping%27%3A%20False%7D%2Cvaccine_schedule%3Dnone%29/reward%3D%28power%3D1.0%2Cscale_factor%3D100%2Chorizon%3D96%29/policy_R.png image

Value table: http://ec2-3-138-203-155.us-east-2.compute.amazonaws.com:8001/results/env%3D%28R_0%3D2.5%2Cdistr_family%3Dpoisson%2Cimported_cases_per_step%3D0.0%2Cdynamics%3DSIS%2Ccustom%3D%7B%27extra_scale%27%3A%201.0%2C%20%27tags%27%3A%20None%2C%20%27time_lumping%27%3A%20False%7D%2Cvaccine_schedule%3Dnone%29/reward%3D%28power%3D1.0%2Cscale_factor%3D100%2Chorizon%3D96%29/value.txt

Old: http://ec2-3-138-203-155.us-east-2.compute.amazonaws.com:8001/results_old/env%3D%28R_0%3D2.5%2Cdistr_family%3Dpoisson%2Cimported_cases_per_step%3D0.0%2Cnum_states%3D101%2Cnum_actions%3D15%2Cdynamics%3DSIS%2Caction_frequency%3D1%2Ccustom%3D%7B%27extra_scale%27%3A%201.0%2C%20%27tags%27%3A%20%27hospital_capacity_limit%27%2C%20%27time_lumping%27%3A%20False%7D%29/reward%3D%28power%3D1.0%2Cscale_factor%3D100%2Chorizon%3D96.0%2Cconvergence_threshold%3D0.0001%2Cdiscount_factor%3D1.0%29/policy.png image

Value table: http://ec2-3-138-203-155.us-east-2.compute.amazonaws.com:8001/results_old/env%3D%28R_0%3D2.5%2Cdistr_family%3Dpoisson%2Cimported_cases_per_step%3D0.0%2Cnum_states%3D101%2Cnum_actions%3D15%2Cdynamics%3DSIS%2Caction_frequency%3D1%2Ccustom%3D%7B%27extra_scale%27%3A%201.0%2C%20%27tags%27%3A%20%27hospital_capacity_limit%27%2C%20%27time_lumping%27%3A%20False%7D%29/reward%3D%28power%3D1.0%2Cscale_factor%3D100%2Chorizon%3D96.0%2Cconvergence_threshold%3D0.0001%2Cdiscount_factor%3D1.0%29/value.txt

blakeelias commented 3 years ago

Value functions are clearly different.