Anatomy of a Request

The default qiotoolkit input format is JSON, although it would be possible to handle other formats with different input handlers.

A minimal request for qiotoolkit has the following shape:

{
  "target": "SOLVER_IDENTIFIER",
  "input_params": {
     ...
  }
}

  "cost_function": {
    "type": "MODEL_IDENTIFIER",
    "version": "MODEL_VERSION",
    ...
  }
}

Where the SOLVER_IDENTIFIER must exactly match the value returned by AbstractSolver::get_identifier().

Likewise, MODEL_IDENTIFIER and MODEL_VERSION must match those return by markov::Model::get_identifier() and markov::Model::get_version().

Note

These identifiers and version numbers not only tell qiotoolkit what to simulate but ensure there is no mismatch in versioning (i.e., you're trying to run old input with a substantially modified newer version of the code or vice-versa.

Input Parameters

Additional configuration for the solver should be denoted in the input_params section of the request. While some parameters are shared between all solvers (e.g., time_limit), others are specific to the solver. For instance, the Parallel Tempering solver can be configured with:

  ...
  "input_params": {
    "max_steps": 10,
    "seed": 42,
    "temperature_set": {
      "type": "geometric",
      "low": 0.5,
      "high": 3,
      "steps": 32
    }
  },
  ...

Note

For the exact set of parameters understood by a solver, refer to the solver specifications.

Input Data

The model is configured from additional parameters in the input_data section. These input parameters determine which cost function will be optimized. For instance, the Ising model expects a term entry describing the terms of the polynomial in the Ising cost function:

  ...
  "input_data": {
    "type": "ising",
    "version": "1.0",
    "terms": [
      {"c": -1, "ids": [0,1]},
      {"c": 2, "ids": [0,2]},
      {"c": 1, "ids": [1,2]},
      ...
    ]
  },
  ...

Which, in this case, describes a cost function of the form

  \mathcal{H} = -s_0s_1 + 2s_0s_2 + s_1s_2 + \cdots

Note

For the exact set of parameters required by each model, refer to the model specifications.