Simulated Annealing

Performs metropolis dynamics on the model while gratually lowering ("annealing") the simulation temperature. As a result, the system settles in a low-cost state while still being allowed to escape local minima as the simulation progresses.

qiotoolkit's simulated annealing implementation allows the specification of an annealing schedule either as an explicit array of values or as a schedule-generator.
This (explicit or schedule-generated) list of numbers can be interpreted as inverse temperatures by specifying use_inverse_temperature = true.
Multiple restarts can be performed using the restarts parameter. These restarts are mutually independent.

Note

If you use the QIOtoolkit-comptaible parameters beta_start and beta_stop, the solver will attempt to infer the number of annealing steps from step_limit or its Qiotoolkit-alias sweeps.

Example

{
  "target": "simulated-annealing.qiotoolkit",
  "version": "1.0",
  "input_params": {
    "seed": 42,
    "schedule": {
      "type": "geometric",
      "initial_value": 3.0,
      "final_value": 0.2,
      "count": 128
    }
    "restarts": 8
  },
  "input_data_uri": "..."
}

This will perform 8 restarts of cooling the system from T=3.0 to T=0.2 in 128 steps spaced geometrically.

param	type	required/default	description
`seed`	integer	default: time-based
`schedule`	Schedule required	How the temperature should be changed over time.
`use_inverse_temperature`	boolean	default: false	Whether to interpret `schedule` as inverse temperature.
`beta_start`	float >0	(QIO compatibility)	Float specifying the starting inverse temperature.
`beta_stop`	float >beta_start	(QIO compatibility)	Float specifying the stopping inverse temperature.
`restarts`	integer > 0	default: 1	How many restarts to perform.

{
  "$id": "solver/simulatedannealing.qiotoolkit.schema",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "Simulated Annealing on CPU",
  "type": "object",
  "required": ["target", "input_params"],
  "properties": {
    "target": {
      "type": "string",
      "const": "simulatedannealing.qiotoolkit"
    },
    "input_params": {
      "type": "object",
      "oneOf": [
        {"requried": ["schedule"]},
        {"required": ["beta_start", "beta_stop"]},
      ],
      "properties": {
        "schedule": {
          "type": "object",
          "$ref": "../schedule/schedule.schema",
          "description": "How the temperature should be changed over time."
        },
        "use_inverse_temperature": {
          "type": "bool",
          "description": "Whether to interpret `schedule` as inverse temperature",
          "default": false
        },
        "beta_start": {
          "type": "number",
          "description": "float specifying the starting inverse temperature."
          "minimum": 0,
        },
        "beta_stop": {
          "type": "number",
          "description": "float specifying the stopping inverse temperature."
          "minimum": 0,
        },
        "restarts": {
          "type": "number",
          "description": "how many restarts to perform",
          "minimum": 1,
          "multipleOf": 1,
          "default": 1
        }
      },
    }
  }
}