package tensorflow.decision_trees

Get desktop application:
View/edit binary Protocol Buffers messages

When present, the Ensemble's output is the average of member models' outputs.

Used in: Ensemble

• repeated google.protobuf.Any additional_data = 1

Used in: TreeNode, tensorforest.SplitCandidate

• optional google.protobuf.Int32Value left_child_id = 1

• optional google.protobuf.Int32Value right_child_id = 2

• BinaryNode.Direction default_direction = 3

  When left_child_test is undefined for a particular datapoint (e.g. because it's not defined when feature value is missing), the datapoint should go in this direction.

• oneof left_child_test

  When a datapoint satisfies the test, it should be propagated to the left child.

  • InequalityTest inequality_left_child_test = 4

  • google.protobuf.Any custom_left_child_test = 5

Used in: BinaryNode

• LEFT = 0

• RIGHT = 1

Used in: Model

• repeated TreeNode nodes = 1

• repeated google.protobuf.Any additional_data = 2

An ordered sequence of models. This message can be used to express bagged or boosted models, as well as custom ensembles.

Used in: Model

• repeated Ensemble.Member members = 100

  A higher id for more readable printing.

• oneof combination_technique

  The presence of a certain combination_technique indicates how to combine the outputs of member models in order to compute the ensemble's output.

  • Summation summation_combination_technique = 1

  • Averaging averaging_combination_technique = 2

  • google.protobuf.Any custom_combination_technique = 3

• repeated google.protobuf.Any additional_data = 4

Used in: Ensemble

• optional Model submodel = 1

• optional google.protobuf.Int32Value submodel_id = 2

• repeated google.protobuf.Any additional_data = 3

Used in: InequalityTest, MatchingValuesTest, ModelAndFeatures.Feature, ObliqueFeatures

• optional google.protobuf.StringValue id = 1

• repeated google.protobuf.Any additional_data = 2

Used in: BinaryNode

• oneof FeatureSum

  When the feature is missing, the test's outcome is undefined.

  • FeatureId feature_id = 1

  • ObliqueFeatures oblique = 4

• InequalityTest.Type type = 2

• optional Value threshold = 3

Used in: InequalityTest, tensorforest.TensorForestParams

• LESS_OR_EQUAL = 0

• LESS_THAN = 1

• GREATER_OR_EQUAL = 2

• GREATER_THAN = 3

Used in: TreeNode

• oneof leaf

  • Vector vector = 1

    The interpretation of the values held in the leaves of a decision tree is application specific, but some common cases are: 1) len(vector) = 1, and the floating point value[0] holds the class 0 probability in a two class classification problem. 2) len(vector) = 1, and the integer value[0] holds the class prediction. 3) The floating point value[i] holds the class i probability prediction. 4) The floating point value[i] holds the i-th component of the vector prediction in a regression problem. 5) sparse_vector holds the sparse class predictions for a classification problem with a large number of classes.

  • SparseVector sparse_vector = 2

• repeated google.protobuf.Any additional_data = 3

  For non-standard handling of leaves.

Used in generic_tree_model.BinaryNode.left_child_test. Tests whether the feature's value belongs to the specified list, (or does not belong if inverse=True).

• optional FeatureId feature_id = 1

  When the feature is missing, the test's outcome is undefined.

• repeated Value value = 2

• bool inverse = 3

A generic handle for any type of model.

Used in: Ensemble.Member, ModelAndFeatures

• oneof model

  • DecisionTree decision_tree = 1

  • Ensemble ensemble = 2

  • google.protobuf.Any custom_model = 3

• repeated google.protobuf.Any additional_data = 4

• map<string, ModelAndFeatures.Feature> features = 1

  Given a FeatureId feature_id, the feature's description is in features[feature_id.id.value].

• optional Model model = 2

• repeated google.protobuf.Any additional_data = 3

Used in: ModelAndFeatures

• optional FeatureId feature_id = 1

  TODO(jonasz): Remove this field, as it's confusing. Ctx: cr/153569450.

• repeated google.protobuf.Any additional_data = 2

Used in: InequalityTest

• repeated FeatureId features = 1

  total value is sum(features[i] * weights[i]).

• repeated float weights = 2

A SparseVector represents a vector in which only certain select elements are non-zero. Maps labels to values (e.g. class id to probability or count).

Used in: Leaf, tensorforest.LeafStat.GiniImpurityClassificationStats

• map<int64, Value> sparse_value = 1

When present, the Ensemble's output is the sum of member models' outputs.

Used in: Ensemble

• repeated google.protobuf.Any additional_data = 1

Used in: DecisionTree

• optional google.protobuf.Int32Value node_id = 1

  Following fields are provided for convenience and better readability. Filling them in is not required.

• optional google.protobuf.Int32Value depth = 2

• optional google.protobuf.Int32Value subtree_size = 3

• oneof node_type

  • BinaryNode binary_node = 4

  • Leaf leaf = 5

  • google.protobuf.Any custom_node_type = 6

• repeated google.protobuf.Any additional_data = 7

Represents a single value of any type, e.g. 5 or "abc".

Used in: InequalityTest, MatchingValuesTest, SparseVector, Vector

• oneof value

  • float float_value = 1

  • double double_value = 2

  • int32 int32_value = 3

  • int64 int64_value = 4

  • google.protobuf.Any custom_value = 5

Used in: Leaf, tensorforest.LeafStat.GiniImpurityClassificationStats, tensorforest.LeafStat.LeastSquaresRegressionStats

• repeated Value value = 1