

Structured Data for Machine Learning Model

Below, the user can find an example JSON structured representation for the Machine Learning Model.

Schema

{
    "description": "machine learning model type/subtype schema", 
    "allOf": [
        {
            "schemaId": "model", 
            "required": [
                "tier1", 
                "tier2", 
                "tier3", 
                "type", 
                "method"
            ], 
            "properties": {
                "tier1": {
                    "schemaId": "core-primitive-slugified-entry-or-slug", 
                    "oneOf": [
                        {
                            "description": "container for machine- and human-readable identifier", 
                            "title": "slugified entry", 
                            "schemaId": "core-primitive-slugified-entry", 
                            "required": [
                                "name", 
                                "slug"
                            ], 
                            "type": "object", 
                            "properties": {
                                "name": {
                                    "type": "string", 
                                    "description": "descriptive human-readable name of entry"
                                }, 
                                "slug": {
                                    "type": "string", 
                                    "description": "machine-readable identifier"
                                }
                            }
                        }, 
                        {
                            "type": "string"
                        }
                    ], 
                    "description": "contains either object with slugified entry or slug only as a string", 
                    "title": "slugified entry or slug"
                }, 
                "tier3": {
                    "schemaId": "core-primitive-slugified-entry-or-slug", 
                    "oneOf": [
                        {
                            "description": "container for machine- and human-readable identifier", 
                            "title": "slugified entry", 
                            "schemaId": "core-primitive-slugified-entry", 
                            "required": [
                                "name", 
                                "slug"
                            ], 
                            "type": "object", 
                            "properties": {
                                "name": {
                                    "type": "string", 
                                    "description": "descriptive human-readable name of entry"
                                }, 
                                "slug": {
                                    "type": "string", 
                                    "description": "machine-readable identifier"
                                }
                            }
                        }, 
                        {
                            "type": "string"
                        }
                    ], 
                    "description": "contains either object with slugified entry or slug only as a string", 
                    "title": "slugified entry or slug"
                }, 
                "tier2": {
                    "schemaId": "core-primitive-slugified-entry-or-slug", 
                    "oneOf": [
                        {
                            "description": "container for machine- and human-readable identifier", 
                            "title": "slugified entry", 
                            "schemaId": "core-primitive-slugified-entry", 
                            "required": [
                                "name", 
                                "slug"
                            ], 
                            "type": "object", 
                            "properties": {
                                "name": {
                                    "type": "string", 
                                    "description": "descriptive human-readable name of entry"
                                }, 
                                "slug": {
                                    "type": "string", 
                                    "description": "machine-readable identifier"
                                }
                            }
                        }, 
                        {
                            "type": "string"
                        }
                    ], 
                    "description": "contains either object with slugified entry or slug only as a string", 
                    "title": "slugified entry or slug"
                }, 
                "reference": {
                    "schemaId": "core-reference-literature", 
                    "properties": {
                        "publisher": {
                            "type": "string", 
                            "description": "Publisher of the work."
                        }, 
                        "doi": {
                            "type": "string", 
                            "description": "Digital Object Identifier of the reference."
                        }, 
                        "isbn": {
                            "type": "string", 
                            "description": "International Standard Book Number of the reference."
                        }, 
                        "reference": {
                            "items": {
                                "type": "object"
                            }, 
                            "type": "array", 
                            "description": "References cited by the work. Reference objects can nest as deeply as needed. This is useful, for example, when tracking the history of a value referenced in a scholarly article; the top level reference would contain information about where the data was accessed while the nested reference would contain information about where it was originally published."
                        }, 
                        "title": {
                            "type": "string", 
                            "description": "Title of the work."
                        }, 
                        "url": {
                            "type": "string", 
                            "description": "Internet address of the reference."
                        }, 
                        "issue": {
                            "type": "string", 
                            "description": "Issue of the collection in which the work appeared."
                        }, 
                        "journal": {
                            "type": "string", 
                            "description": "Journal in which the work appeared."
                        }, 
                        "issn": {
                            "type": "string", 
                            "description": "International Standard Serial Number of the reference."
                        }, 
                        "editors": {
                            "items": {
                                "schemaId": "core-reference-literature-name", 
                                "required": [
                                    "first", 
                                    "last"
                                ], 
                                "type": "object", 
                                "properties": {
                                    "affiliation": {
                                        "type": "string"
                                    }, 
                                    "middle": {
                                        "type": "string"
                                    }, 
                                    "last": {
                                        "type": "string"
                                    }, 
                                    "first": {
                                        "type": "string"
                                    }
                                }, 
                                "title": "experiment author schema"
                            }, 
                            "type": "array", 
                            "description": "List of editors of the work."
                        }, 
                        "volume": {
                            "type": "string", 
                            "description": "Volume of the series in which the work appeared."
                        }, 
                        "authors": {
                            "items": {
                                "schemaId": "core-reference-literature-name", 
                                "required": [
                                    "first", 
                                    "last"
                                ], 
                                "type": "object", 
                                "properties": {
                                    "affiliation": {
                                        "type": "string"
                                    }, 
                                    "middle": {
                                        "type": "string"
                                    }, 
                                    "last": {
                                        "type": "string"
                                    }, 
                                    "first": {
                                        "type": "string"
                                    }
                                }, 
                                "title": "experiment author schema"
                            }, 
                            "type": "array", 
                            "description": "List of authors of the work."
                        }, 
                        "year": {
                            "type": "string", 
                            "description": "Year in which the reference was published."
                        }, 
                        "type": {
                            "enum": [
                                "literature"
                            ]
                        }, 
                        "pages": {
                            "type": "object", 
                            "description": "Start and end pages of the work.", 
                            "allOf": [
                                {
                                    "schemaId": "core-reference-literature-pages", 
                                    "required": [
                                        "start"
                                    ], 
                                    "type": "object", 
                                    "properties": {
                                        "start": {
                                            "type": "string"
                                        }, 
                                        "end": {
                                            "type": "string"
                                        }
                                    }, 
                                    "title": "pages schema"
                                }
                            ]
                        }
                    }, 
                    "title": "literature reference schema"
                }, 
                "tags": {
                    "schemaId": "model-tags", 
                    "allOf": [
                        {
                            "items": {
                                "uniqueItems": true, 
                                "type": "string"
                            }, 
                            "schemaId": "core-primitive-array-of-strings", 
                            "type": "array", 
                            "description": "array of strings, e.g. metadata tags", 
                            "title": "array of strings"
                        }
                    ], 
                    "description": "Descriptive (searchable) attributes of model, generic. Example: user-adjustable.", 
                    "title": "model tags schema"
                }, 
                "subtype": {
                    "schemaId": "core-primitive-slugified-entry-or-slug", 
                    "oneOf": [
                        {
                            "description": "container for machine- and human-readable identifier", 
                            "title": "slugified entry", 
                            "schemaId": "core-primitive-slugified-entry", 
                            "required": [
                                "name", 
                                "slug"
                            ], 
                            "type": "object", 
                            "properties": {
                                "name": {
                                    "type": "string", 
                                    "description": "descriptive human-readable name of entry"
                                }, 
                                "slug": {
                                    "type": "string", 
                                    "description": "machine-readable identifier"
                                }
                            }
                        }, 
                        {
                            "type": "string"
                        }
                    ], 
                    "description": "contains either object with slugified entry or slug only as a string", 
                    "title": "slugified entry or slug"
                }, 
                "modifiers": {
                    "schemaId": "model-modifiers", 
                    "allOf": [
                        {
                            "items": {
                                "uniqueItems": true, 
                                "type": "string"
                            }, 
                            "schemaId": "core-primitive-array-of-strings", 
                            "type": "array", 
                            "description": "array of strings, e.g. metadata tags", 
                            "title": "array of strings"
                        }
                    ], 
                    "description": "Additions to a model, that expand upon the underlying physical principle without fundamentally changing the working equations (e.g. in a linear fashion). Examples: Hubbard U, external potential, explicit correlation", 
                    "title": "model modifications schema"
                }, 
                "augmentations": {
                    "schemaId": "model-augmentations", 
                    "allOf": [
                        {
                            "items": {
                                "uniqueItems": true, 
                                "type": "string"
                            }, 
                            "schemaId": "core-primitive-array-of-strings", 
                            "type": "array", 
                            "description": "array of strings, e.g. metadata tags", 
                            "title": "array of strings"
                        }
                    ], 
                    "description": "Descriptive properties of a model, that do not change the underlying physical principles of the model. Example: localization for local-orbital DFT.", 
                    "title": "model augmentation schema"
                }, 
                "type": {
                    "schemaId": "core-primitive-slugified-entry-or-slug", 
                    "oneOf": [
                        {
                            "description": "container for machine- and human-readable identifier", 
                            "title": "slugified entry", 
                            "schemaId": "core-primitive-slugified-entry", 
                            "required": [
                                "name", 
                                "slug"
                            ], 
                            "type": "object", 
                            "properties": {
                                "name": {
                                    "type": "string", 
                                    "description": "descriptive human-readable name of entry"
                                }, 
                                "slug": {
                                    "type": "string", 
                                    "description": "machine-readable identifier"
                                }
                            }
                        }, 
                        {
                            "type": "string"
                        }
                    ], 
                    "description": "contains either object with slugified entry or slug only as a string", 
                    "title": "slugified entry or slug"
                }, 
                "method": {
                    "schemaId": "method", 
                    "required": [
                        "type", 
                        "subtype"
                    ], 
                    "properties": {
                        "subtype": {
                            "type": "string", 
                            "description": "general subtype of this method, eg. `ultra-soft`"
                        }, 
                        "type": {
                            "type": "string", 
                            "description": "general type of this method, eg. `pseudopotential`"
                        }, 
                        "precision": {
                            "type": "object", 
                            "description": "Object showing the actual possible precision based on theory and implementation"
                        }, 
                        "data": {
                            "type": "object", 
                            "description": "additional data specific to method, eg. array of pseudopotentials"
                        }
                    }, 
                    "title": "method schema (base)"
                }
            }, 
            "title": "model schema (base)"
        }
    ], 
    "schemaId": "models-directory-st-det-ml", 
    "title": "machine learning model schema", 
    "$schema": "http://json-schema.org/draft-04/schema#", 
    "properties": {
        "tier3": {
            "enum": [
                {
                    "name": "machine learning", 
                    "slug": "ml"
                }
            ]
        }, 
        "type": {
            "enum": [
                {
                    "name": "regression", 
                    "slug": "re"
                }
            ]
        }, 
        "method": {
            "oneOf": [
                {
                    "required": [
                        "precision", 
                        "data"
                    ], 
                    "schemaId": "methods-directory-regression", 
                    "allOf": [
                        {
                            "schemaId": "method", 
                            "required": [
                                "type", 
                                "subtype"
                            ], 
                            "properties": {
                                "subtype": {
                                    "type": "string", 
                                    "description": "general subtype of this method, eg. `ultra-soft`"
                                }, 
                                "type": {
                                    "type": "string", 
                                    "description": "general type of this method, eg. `pseudopotential`"
                                }, 
                                "precision": {
                                    "type": "object", 
                                    "description": "Object showing the actual possible precision based on theory and implementation"
                                }, 
                                "data": {
                                    "type": "object", 
                                    "description": "additional data specific to method, eg. array of pseudopotentials"
                                }
                            }, 
                            "title": "method schema (base)"
                        }
                    ], 
                    "properties": {
                        "subtype": {
                            "enum": [
                                "least_squares", 
                                "ridge"
                            ]
                        }, 
                        "type": {
                            "enum": [
                                "linear", 
                                "kernel_ridge"
                            ]
                        }, 
                        "precision": {
                            "perProperty": {
                                "items": {
                                    "type": "object", 
                                    "properties": {
                                        "name": {
                                            "type": "string", 
                                            "description": "property name in 'flattened' format"
                                        }
                                    }, 
                                    "allOf": [
                                        {
                                            "schemaId": "methods-directory-regression-precision", 
                                            "oneOf": [
                                                {
                                                    "schemaId": "methods-directory-regression-precision-per-property", 
                                                    "required": [
                                                        "trainingError"
                                                    ], 
                                                    "properties": {
                                                        "score": {
                                                            "type": "number", 
                                                            "description": "prediction score of the estimator. Eg: r2_score"
                                                        }, 
                                                        "trainingError": {
                                                            "type": "number", 
                                                            "description": "training error of the estimator"
                                                        }
                                                    }, 
                                                    "title": "precision schema for regression"
                                                }
                                            ], 
                                            "title": "regression precision"
                                        }
                                    ]
                                }
                            }
                        }, 
                        "data": {
                            "perProperty": {
                                "items": {
                                    "type": "object", 
                                    "properties": {
                                        "name": {
                                            "type": "string", 
                                            "description": "property name in 'flattened' format"
                                        }
                                    }, 
                                    "allOf": [
                                        {
                                            "schemaId": "methods-directory-regression-data", 
                                            "oneOf": [
                                                {
                                                    "schemaId": "methods-directory-regression-linear-data-per-property", 
                                                    "required": [
                                                        "intercept", 
                                                        "perFeature"
                                                    ], 
                                                    "properties": {
                                                        "perFeature": {
                                                            "items": {
                                                                "required": [
                                                                    "coefficient"
                                                                ], 
                                                                "type": "object", 
                                                                "properties": {
                                                                    "coefficient": {
                                                                        "type": "number", 
                                                                        "description": "coefficient in linear regression"
                                                                    }
                                                                }, 
                                                                "allOf": [
                                                                    {
                                                                        "schemaId": "methods-directory-regression-data-per-feature", 
                                                                        "required": [
                                                                            "name"
                                                                        ], 
                                                                        "properties": {
                                                                            "importance": {
                                                                                "type": "number", 
                                                                                "description": "pvalue: https://en.wikipedia.org/wiki/P-value"
                                                                            }, 
                                                                            "name": {
                                                                                "type": "string", 
                                                                                "description": "feature name"
                                                                            }
                                                                        }, 
                                                                        "title": "feature parameters schema"
                                                                    }
                                                                ]
                                                            }, 
                                                            "type": "array", 
                                                            "description": "per-feature (property used for training the ML method/model) parameters"
                                                        }, 
                                                        "intercept": {
                                                            "type": "number", 
                                                            "description": "intercept (shift) from the linear or non-linear fit of data points"
                                                        }
                                                    }, 
                                                    "title": "linear regression parameters schema"
                                                }, 
                                                {
                                                    "schemaId": "methods-directory-regression-kernel-ridge-data-per-property", 
                                                    "required": [
                                                        "xFit", 
                                                        "dualCoefficients", 
                                                        "perFeature"
                                                    ], 
                                                    "properties": {
                                                        "dualCoefficients": {
                                                            "type": "array", 
                                                            "description": "dual coefficients"
                                                        }, 
                                                        "xFit": {
                                                            "type": "array", 
                                                            "description": "training data"
                                                        }, 
                                                        "perFeature": {
                                                            "items": {
                                                                "required": [
                                                                    "name"
                                                                ], 
                                                                "type": "object", 
                                                                "allOf": [
                                                                    {
                                                                        "schemaId": "methods-directory-regression-data-per-feature", 
                                                                        "required": [
                                                                            "name"
                                                                        ], 
                                                                        "properties": {
                                                                            "importance": {
                                                                                "type": "number", 
                                                                                "description": "pvalue: https://en.wikipedia.org/wiki/P-value"
                                                                            }, 
                                                                            "name": {
                                                                                "type": "string", 
                                                                                "description": "feature name"
                                                                            }
                                                                        }, 
                                                                        "title": "feature parameters schema"
                                                                    }
                                                                ]
                                                            }, 
                                                            "type": "array", 
                                                            "description": "per-feature (property used for training the ML method/model) parameters"
                                                        }
                                                    }, 
                                                    "title": "linear regression parameters schema"
                                                }
                                            ], 
                                            "title": "regression data"
                                        }
                                    ]
                                }
                            }, 
                            "dataSet": {
                                "schemaId": "methods-directory-regression-dataset", 
                                "required": [
                                    "exabyteIds"
                                ], 
                                "type": "object", 
                                "description": "dataset for ml", 
                                "properties": {
                                    "exabyteIds": {
                                        "items": {
                                            "type": "string"
                                        }, 
                                        "type": "array", 
                                        "description": "array of exabyteIds for materials in dataset"
                                    }, 
                                    "extra": {
                                        "description": "holder for any extra information, eg. coming from user-uploaded CSV file"
                                    }
                                }
                            }
                        }
                    }, 
                    "title": "linear methods schema"
                }
            ]
        }
    }
}

Example

{
    "tier1": {
        "name": "statistical", 
        "slug": "st"
    }, 
    "tier3": {
        "name": "machine learning", 
        "slug": "ml"
    }, 
    "tier2": {
        "name": "deterministic", 
        "slug": "det"
    }, 
    "type": {
        "name": "regression", 
        "slug": "re"
    }, 
    "method": {
        "subtype": "least_squares", 
        "data": {
            "perProperty": [
                {
                    "perFeature": [
                        {
                            "coefficient": 0.015, 
                            "importance": 0.134, 
                            "name": "atomic_radius:Ge"
                        }, 
                        {
                            "coefficient": 0.016, 
                            "importance": 0.135, 
                            "name": "atomic_radius:Si"
                        }
                    ], 
                    "intercept": 0.363, 
                    "name": "band_gaps:direct"
                }, 
                {
                    "perFeature": [
                        {
                            "coefficient": 0.016, 
                            "importance": 0.135, 
                            "name": "atomic_radius:Ge"
                        }, 
                        {
                            "coefficient": 0.017, 
                            "importance": 0.136, 
                            "name": "atomic_radius:Si"
                        }
                    ], 
                    "intercept": 0.364, 
                    "name": "band_gaps:indirect"
                }
            ], 
            "dataSet": {
                "exabyteIds": [
                    "LCthJ6E2QabYCZqf4", 
                    "LCthJ6E2QabYCZqf5", 
                    "LCthJ6E2QabYCZqf6", 
                    "LCthJ6E2QabYCZqf7", 
                    "LCthJ6E2QabYCZqf8", 
                    "LCthJ6E2QabYCZqf9", 
                    "LCthJ6E2QabYCZq10", 
                    "LCthJ6E2QabYCZq11"
                ], 
                "extra": {}
            }
        }, 
        "precision": {
            "perProperty": {
                "score": 0.8, 
                "trainingError": 0.002
            }
        }, 
        "type": "linear"
    }
}