Pyciemss optimize multiple constraints (#117)

* Step 1

* correction to qoi + riskbound

* more info in QoI as isMinimized effects more

* rename to qoi_methods as its a bit better

* correct test request

* correction to optimzie test request

service/models/operations/optimize.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import ClassVar, List, Optional
+from typing import ClassVar, List, Optional, Union
 from enum import Enum
 from utils.rabbitmq import OptimizeHook
 from pika.exceptions import AMQPConnectionError
@@ -31,17 +31,33 @@ class QOIMethod(str, Enum):
 class QOI(BaseModel):
  method: QOIMethod = QOIMethod.day_average
  contexts: List[str] = []
+ risk_bound: float = 0
+ is_minimized: bool = True
 
  def gen_call(self):
  contexts = self.contexts
- qoi_map = {
- QOIMethod.day_average: lambda samples: obs_nday_average_qoi(
- samples, contexts, 1
- ),
- QOIMethod.max: lambda samples: obs_max_qoi(samples, contexts),
- }
+ if self.is_minimized is True:
+ qoi_map = {
+ QOIMethod.day_average: lambda samples: obs_nday_average_qoi(
+ samples, contexts, 1
+ ),
+ QOIMethod.max: lambda samples: obs_max_qoi(samples, contexts),
+ }
+ else:
+ qoi_map = {
+ QOIMethod.day_average: lambda samples: -obs_nday_average_qoi(
+ samples, contexts, 1
+ ),
+ QOIMethod.max: lambda samples: -obs_max_qoi(samples, contexts),
+ }
  return qoi_map[self.method]
 
+ def gen_risk_bound(self):
+ if self.is_minimized is True:
+ return self.risk_bound
+ else:
+ return -self.risk_bound
+
 
 def objfun(x, initial_guess, objective_function_option):
  if objective_function_option == "lower_bound":
@@ -80,7 +96,7 @@ class OptimizeExtra(BaseModel):
  )
  maxiter: int = 5
  maxfeval: int = 25
- alpha: float = 0.95
+ alpha: Union[List[float], float] = 0.95
  solver_method: str = "dopri5"
  # https://github.com/ciemss/pyciemss/blob/main/pyciemss/integration_utils/interface_checks.py
  solver_step_size: float = Field(
@@ -99,8 +115,7 @@ class Optimize(OperationRequest):
  None
  ) # Theses are interventions provided that will not be optimized
  logging_step_size: float = 1.0
- qoi: QOI
- risk_bound: float
+ qoi: list[QOI]
  bounds_interventions: List[List[float]]
  extra: OptimizeExtra = Field(
  None,
@@ -170,6 +185,12 @@ def gen_pyciemss_args(self, job_id):
  def progress_hook(current_results):
  logging.info(f"Optimize current results: {current_results.tolist()}")
 
+ qoi_methods = []
+ risk_bounds = []
+ for qoi in self.qoi:
+ qoi_methods.append(qoi.gen_call())
+ risk_bounds.append(qoi.gen_risk_bound())
+
  return {
  "model_path_or_json": amr_path,
  "logging_step_size": self.logging_step_size,
@@ -180,8 +201,8 @@ def progress_hook(current_results):
  self.optimize_interventions.initial_guess[0],
  self.optimize_interventions.objective_function_option[0],
  ),
- "qoi": self.qoi.gen_call(),
- "risk_bound": self.risk_bound,
+ "qoi": qoi_methods,
+ "risk_bound": risk_bounds,
  "initial_guess_interventions": self.optimize_interventions.initial_guess,
  "bounds_interventions": self.bounds_interventions,
  "static_parameter_interventions": optimize_interventions,

tests/examples/optimize/input/request.json

@@ -14,14 +14,14 @@
  "start": 0,
  "end": 90
  },
- "qoi": {
+ "qoi": [{
  "contexts": ["Infected_state"],
- "method": "day_average"
- },
- "risk_bound": 10.0,
+ "method": "day_average",
+ "risk_bound": 10.0,
+ "is_minimized": true
+ }],
  "bounds_interventions": [[0.0], [3.0]],
  "extra": {
- "num_samples": 4,
- "is_minimized": true
+ "num_samples": 4
  }
 }