Feature/sof 2002

example/job/ml_model.json

@@ -0,0 +1,11 @@
+{
+    "name": "model",
+    "type": "ml",
+    "subtype": "supervised",
+    "references": [
+        "http://scikit-learn.org/stable/tutorial/machine_learning_map/index.html"
+    ],
+    "method": {
+        "...": "include(model/ml_method.json)"
+    }
+}

example/job/model/method/ml_workflow.json

@@ -0,0 +1,14 @@
+{
+    "units": [
+        {
+            "...": "include(workflow/ml_interface.json)"
+        },
+        {
+            "...": "include(workflow/ml_fit.json)"
+        },
+        {
+            "...": "include(workflow/ml_predict.json)"
+        }
+    ],
+    "name": "property-prediction"
+}

example/job/model/method/workflow/ml_fit.json

@@ -0,0 +1,11 @@
+
+{
+    "head": false,
+    "flowchartId": "2",
+    "name": "ModelFit",
+    "type": "fit",
+    "fit": {
+        "...": "include(unit/fit.json)"
+    }
+
+}

example/job/model/method/workflow/ml_interface.json

@@ -0,0 +1,9 @@
+{
+    "head": true,
+    "flowchartId": "1",
+    "name": "MLInterface",
+    "type": "interface",
+    "interface": {
+        "...": "include(unit/interface.json)"
+    }
+}

example/job/model/method/workflow/ml_predict.json

@@ -0,0 +1,9 @@
+{
+    "head": false,
+    "flowchartId": "3",
+    "name": "Predict",
+    "type": "predict",
+    "predict": {
+        "...": "include(unit/predict.json)"
+    }
+}

example/job/model/method/workflow/unit/fit.json

@@ -0,0 +1,14 @@
+{
+    "compute": null,
+    "input": [
+        {
+            "algorithm": "linear_regression",
+            "fitParameters": [
+                "sample_weight"
+                ],
+            "targetProperties":[
+                "band_gaps"
+            ]
+        }
+    ]
+}

example/job/model/method/workflow/unit/interface.json

@@ -0,0 +1,27 @@
+{
+    "compute": null,
+    "input": [
+        {
+            "chemicalSystem": [
+                "Si",
+                "Ge"
+            ],
+            "superset": false,
+            "databases": [
+                "ExabyteDB"
+                ],
+            "targetMaterials":[
+                "Si5Ge11"
+            ],
+            "features":[
+                "Si",
+                "Ge",
+                "density"
+            ],
+            "targetProperties":[
+                "band_gaps"
+            ],
+            "owner" : "paxJukXpDBiSRjfRZ"
+        }
+    ]
+}

example/job/model/method/workflow/unit/predict.json

@@ -0,0 +1,10 @@
+{
+    "compute": null,
+    "input": [
+        {
+            "targetProperties":[
+                "band_gaps"
+            ]
+        }
+    ]
+}

example/job/model/ml_method.json

@@ -0,0 +1,7 @@
+{
+    "type": "prediction",
+    "subtype": "linear_regression",
+    "workflow": {
+        "...": "include(method/ml_workflow.json)"
+    }
+}

example/job/theory/ml/model_parameters/input_data

@@ -0,0 +1 @@
+"{\"Si\":{\"c8zG3XcnnTybfP3QQ\":0.6875,\"wWK2XCyivQrG8j34Z\":0.8125},\"Ge\":{\"c8zG3XcnnTybfP3QQ\":0.3125,\"wWK2XCyivQrG8j34Z\":0.1875},\"lattice_volume\":{\"c8zG3XcnnTybfP3QQ\":327.1127749111,\"wWK2XCyivQrG8j34Z\":327.1127749111}}"

example/job/theory/ml/model_parameters/model_parameters.json

@@ -0,0 +1,18 @@
+{
+    "algorithm": "linear_regression",
+    "task": "property_prediction",
+    "property": "band_gaps",
+    "modelParameters":{
+        "modelCoefficients":[
+            -3.339e-01,
+            2.555e-01,
+            -1.005e-02
+        ],
+        "intercept": 0.45,
+        "score": 0.8,
+        "modelData" : {
+            "...": "(include(input_data)"
+        }
+
+    }
+}

example/job/theory/ml/prediction.json

@@ -0,0 +1,10 @@
+{
+    "type": "prediction",
+    "subtype":"linear_regression",
+    "precision":{
+        "...": "include(prediction/precision.json)"
+    },
+    "data":{
+        "...": "include(model_parameters/model_parameters.json)"
+    }
+}

example/job/theory/ml/prediction/precision.json

@@ -0,0 +1,3 @@
+{
+    "score": 0.8
+}

schema/job/model.json

@@ -27,6 +27,9 @@
     "oneOf": [
         {
             "$ref": "file:theory/dft.json"
+        },
+        {
+            "$ref": "file:theory/ml.json"
         }
     ],
     "required": [

schema/job/model/method.json

@@ -28,6 +28,12 @@
         },
         {
             "$ref": "file:../theory/dft/local-orbital.json"
+        },
+        {
+            "$ref": "file:../theory/ml/prediction.json"
+        },
+        {
+            "$ref": "file:../theory/ml/classification.json"
         }
     ],
     "required": [

schema/job/model/method/workflow/unit.json

@@ -12,7 +12,11 @@
                 "reduce",
                 "assignment",
                 "condition",
-                "exit"
+                "exit",
+                "interface",
+                "fit",
+                "predict",
+                "transform"
             ]
         },
         "name": {
@@ -68,6 +72,10 @@
                 "$ref": "file:unit/_result.json"
             }
         },
+        "modelParameters":{
+            "description": "parameters from Machine Learning model fitting",
+            "$ref": "file:../../../theory/ml/model_parameters/model_parameters.json"
+        },
         "execution": {
             "description": "Contains information about the execution unit..",
             "$ref": "file:unit/execution.json"
@@ -83,6 +91,22 @@
         "map": {
             "description": "Contains information about the map unit.",
             "$ref": "file:unit/map.json"
+        },
+        "interface": {
+            "description": "Contains information about the Machine Learning interface unit.",
+            "$ref": "file:unit/interface.json"
+        },
+        "fit": {
+            "description": "Contains information about the Machine Learning model fit unit.",
+            "$ref": "file:unit/fit.json"
+        },
+        "predict": {
+            "description": "Contains information about the Machine Learning property predict unit.",
+            "$ref": "file:unit/predict.json"
+        },
+        "transform": {
+            "description": "Contains information about the Machine Learning data transform unit.",
+            "$ref": "file:unit/transform.json"
         }
     },
     "required": [

schema/job/model/method/workflow/unit/fit.json

@@ -0,0 +1,35 @@
+{
+    "$schema": "http://json-schema.org/draft-04/schema#",
+    "title": "ML Fit unit schema",
+    "type": "object",
+    "properties": {
+        "input": {
+            "description": "Unit's input.",
+            "type": "array",
+            "items": {
+                "type": "object",
+                "properties": {
+                    "algorithm": {
+                        "description": "Algorithm name",
+                        "type": "string"
+                    },
+                    "fitParameters": {
+                        "description": "Additional paramaters used for model fitting",
+                        "type": "array"
+                    },
+                    "targetProperties": {
+                        "description": "Target properties to predict",
+                        "type": "array"
+                    }
+                },
+                "required": [
+                    "algorithm",
+                    "targetProperties"
+                ]
+            }
+        }
+    },
+    "required": [
+        "input"
+    ]
+}

schema/job/model/method/workflow/unit/interface.json

@@ -0,0 +1,57 @@
+{
+    "$schema": "http://json-schema.org/draft-04/schema#",
+    "title": "ML Interface unit schema",
+    "type": "object",
+    "properties": {
+        "input": {
+            "description": "Unit's input.",
+            "type": "array",
+            "items": {
+                "type": "object",
+                "properties": {
+                    "trainingMaterials": {
+                        "description": "List of training materials. Eg: ['Si3Ge5', 'SiGe']",
+                        "type": "array"
+                    },
+                    "chemicalSystem": {
+                        "description": "List of elements in chemical system. Eg: ['Si', 'Ge']",
+                        "type": "array"
+                    },
+                    "superset": {
+                        "description": "Input to define the list of materials formed by the elements in chemical system",
+                        "type": "boolean"
+                    },
+                    "databases": {
+                        "description": "List of databases to get data from",
+                        "type": "array"
+                    },
+                    "features": {
+                        "description": "List of features to use in training. Example: descriptive properties",
+                        "type": "array"
+                    },
+                    "targetMaterials": {
+                        "description": "List of target materials. Eg: ['Si3Ge5', 'SiGe']",
+                        "type": "array"
+                    },
+                    "targetProperties": {
+                        "description": "List of target properties",
+                        "type": "array"
+                    },
+                    "owner":{
+                        "description": "Exabyte ID of the material owner",
+                        "type": "string"
+                    }
+                },
+                "required": [
+                    "chemicalSystem",
+                    "databases",
+                    "targetMaterials",
+                    "targetProperties"
+                ]
+            }
+        }
+    },
+    "required": [
+        "input"
+    ]
+}

schema/job/model/method/workflow/unit/predict.json

@@ -0,0 +1,26 @@
+{
+    "$schema": "http://json-schema.org/draft-04/schema#",
+    "title": "ML Predict unit schema",
+    "type": "object",
+    "properties": {
+        "input": {
+            "description": "Unit's input.",
+            "type": "array",
+            "items": {
+                "type": "object",
+                "properties": {
+                    "targetProperties": {
+                        "description": "Target properties to predict",
+                        "type": "array"
+                    }
+                },
+                "required": [
+                    "targetProperties"
+                ]
+            }
+        }
+    },
+    "required": [
+        "input"
+    ]
+}

schema/job/model/method/workflow/unit/transform.json

@@ -0,0 +1,31 @@
+{
+    "$schema": "http://json-schema.org/draft-04/schema#",
+    "title": "ML Transform unit schema",
+    "type": "object",
+    "properties": {
+        "input": {
+            "description": "Unit's input.",
+            "type": "array",
+            "items": {
+                "type": "object",
+                "properties": {
+                    "algorithm": {
+                        "description": "Algorithm name",
+                        "type": "string"
+                    },
+                    "fitParameters": {
+                        "description": "Additional paramaters used for model fitting",
+                        "type": "array"
+                    }
+                },
+                "required": [
+                    "algorithm",
+                    "fitParameters"
+                ]
+            }
+        }
+    },
+    "required": [
+        "input"
+    ]
+}

schema/job/theory/ml.json

@@ -0,0 +1,18 @@
+{
+    "$schema": "http://json-schema.org/draft-04/schema#",
+    "title": "machine learning schema",
+    "description": "machine learning schema",
+    "properties": {
+        "type": {
+            "enum": [
+                "ml"
+            ]
+        },
+        "subtype": {
+            "enum": [
+                "supervised",
+                "unsupervised"
+            ]
+        }
+    }
+}