Numerical Hard SOM model

Representation

Keys _id, name, stepId, domainInformation, dataset, project, processingInfo, creationTS , latestUpdateTS are unchanged and follow the classic of representation only dataSpecification keys are changed and are described as above :

• dataSpecification:
  • keyword: Its value is “numericalHardSOMModel” . Cf to mandatory keys.
  • valueType: Cf to mandatory keys.
    • dataType: Its value is “numerical” . Cf to mandatory keys.
    • structureType: Its value is “vector” . Cf to mandatory keys.
  • meaning: Its value is “Numerical Hard SOM Model” . Cf to mandatory keys.
  • view: Cf to mandatory keys.
  • dataLocationId: Cf to mandatory keys.
  • hyperParameters: Currently defined in representation.
    • gridProperties : Dictionary of grid properties :
      • width: Integer value for grid width.
      • length: Integer value for grid length.
    • tMax: Double which describe maximum temperature from where starts iterations.
    • tMin: Double which describe minimal temperature from where stops iterations.
    • lastT: Double which describe temperature value at last iteration.
    • neighborhood: Grid description, only rectangular is available for the moment.

JSON example :

{
  "dataSpecification": {
    "keyword": "numericalHardSOMModel",
    "valueType": {
      "dataType": "numerical",
      "structureType": "vector"
    },
    "meaning": "Numerical Hard SOM Model",
    "view": {
      "id": "637ce534dd85c10875c4fe26",
      "name": "view_11-22-2022_15:05:24"
    },
    "dataLocationId": "62b18d804ae71c6a0025237a",
    "hyperParameters": {
      "neighborhood": "rectangular",
      "gridProperties": {
        "width": 5,
        "length": 5
      },
      "tMax": 13.3,
      "tMin": 2.2,
      "lastT": 2.7
    }
  }
}   

(numerical_hard_som_model_obs)=

Observation

The observation value key is a dictionary with three keys :

• clusterId : Integer which specify to which cluster this SOM prototype belongs.
• prototype :
  • name: Name of the feature
  • value: $R^n$ vector describing this prototype features.
• clusterSize : Number of learning observations catch at the last iteration of the SOM algorithm.

{
  "_id": "ObjectId('635905813fd9893f20271d2a')",
  "clusterId": 17,
  "clusterSize": 0,
  "prototype": [
    {
      "name": "cp_a_1_bar",
      "value": -0.15913712480113543
    },
    {
      "name": "cp_a_2_bar",
      "value": -0.1617893554732556
    },
    {
      "name": "cp_r_bar",
      "value": -0.18425408275916427
    }
  ],
  "representationId": "ObjectId('635905813fd9893f20271d2a')"
}