Magpie comes equipped with many different kinds of datasets, models, crystal structure prediction algorithms, and other kinds of variables. This section includes all of the currently available variable types and links to pages that describe what operations they support. If you are not yet familiar with how to call these operations, please consult the documentation for the text interface.
Each of these dataset objects can be used to represent different kinds of data, both in terms of how Magpie represents an entry internally and what kind of attributes it can generate.
data.Dataset: Usage: *No options to set* data.MultiPropertyDataset: Usage: *No options* data.materials.CompositionDataset: Usage: *No options* data.materials.CrystalStructureDataset: Usage: *No options* data.materials.ElementDataset: Usage: *No options* data.materials.PrototypeDataset: Usage: <Structure description filename>Magpie is equipped with the ability to generate many different kinds of models. This includes models for classifying data into known subsets or predicting the value of some property.
Classifiers are used decide which group an entry belongs out of a finite list of options.
models.classification.CumulantExpansionClassifier: Usage: <filename> <order> models.classification.ScikitLearnClassifier: Usage: <model> [<compression level>] models.classification.SplitClassifier: Usage: *No options* models.classification.WekaClassifier: Usage: <Weka classifier> [<classifier options...>]Regression models are used to approximate unknown, continuous functions (think y = f(x) = a + b * x).
models.regression.ClassificationRegression: Usage: $<classifier> <threshold> <objective function> [<o.f. options...>] models.regression.CompositeRegression: Usage: *No options to set* models.regression.GuessMeanRegression: Usage: [-jitter <jitter>] models.regression.LASSORegression: Usage: -maxterms <terms> models.regression.LinearCorrectedRegression: Usage: $<submodel> models.regression.MetallurgicalHeuristicRegression: Usage: $<hull data> [-correction] models.regression.MixingRuleRegression: Usage: <property> [-invert] [-correct] [-fit <ElementNames>] models.regression.MultiObjectiveRegression: Usage: <ranker method> [<ranker options...>] models.regression.MultiPropertyRegression: Usage: <properties...> models.regression.NonlinearRegressionExpr: Usage: <equation to be fit...> models.regression.PolynomialRegression: Usage: <Order> [-print_accuracy <figs>] models.regression.RandomGuessRegression: Usage: <lower bound> <upper bound> models.regression.ScikitLearnRegression: Usage: <model> [<compression level>] models.regression.SingleGuessRegression: Usage: <guess> [-jitter <jitter>] models.regression.SplitRegression: Usage: *No options* models.regression.StagedRegression: Usage: <absolute|relative> models.regression.WekaRegression: Usage: <Weka classifier> [<classifier options...>] models.regression.crystal.CoulombEwaldMatrixRegression: Usage: <lambda> <sigma> models.regression.crystal.CoulombSineMatrixRegression: Usage: <lambda> <sigma> models.regression.crystal.PRDFRegression: Usage: <lambda> <sigma> <cutoff> <bins> models.regression.nonlinear.SimpleLinearModelExample: Usage: *No options*Each of these objects can be used calculate different statistics about the performance of a model.
statistics.performance.ClassificationStatistics: Usage: *No options* statistics.performance.RegressionStatistics: Usage: *No options* statistics.performance.TargetRegressionStatistics: Usage: <target> [-accept <Acceptance Tolerance>] [-window <Max Window Size>] [-cands <Max Number Candidates>]Clustering algorithms perform unsupervised learning, which recognizes groups of data with similar attributes and provides rules for how to distinguish between them. These groups are not known beforehand, use classification algorithms to build rules for separating data into already-known groups.
cluster.WekaClusterer: Usage: <method> [<options...>]Crystal structure prediction algorithms are used to predict which crystal structure is most probable out of a list of known prototypes to be stable at a certain composition.
csp.CompositionBasedCSPEngine: Usage: $<model template> csp.DMSPEngine: Usage: *No options*