CreateSolution : algorithmHyperParameters vs algorithmHyperParameterRanges

algorithmHyperparameters allows you to specify exact values for the hyperparameters, whereas algorithmHyperParameterRanges is used during hyperparameter optimization to set the range for the hyperparameters you want to optimize. The example below (taken from this page of the documentation) shows the difference. The hidden_dimension variable is set to an exact value of 55 using algorithmHyperParameters. The bptt variable is set to be an integer value between 20 and 40 using algorithmHyperParameterRanges under hpoConfig - the exact value will be determined during hyperparameter optimization. Note that performHPO must be set to true to activate the optimization process. The example also shows how to handle a boolean variable for optimization through the recency_mask variable, and continuous variables are also supported.

{
  "performAutoML": false,
  "recipeArn": "arn:aws:personalize:::recipe/aws-hrnn",
  "performHPO": true,
  "solutionConfig": {
      "algorithmHyperParameters": {
          "hidden_dimension": "55"
      },
      "hpoConfig": {
          "algorithmHyperParameterRanges": {
              "categoricalHyperParameterRanges": [
                  {
                      "name": "recency_mask",
                      "values": [ "true", "false" ]
                  }
              ],
              "integerHyperParameterRanges": [
                  {
                      "name": "bptt",
                      "minValue": 20,
                      "maxValue": 40
                  }
              ]
          },
          "hpoResourceConfig": {
              "maxNumberOfTrainingJobs": "4",
              "maxParallelTrainingJobs": "2"
          }
      }
  }
}