Package org.tribuo.math

Interface FeedForwardParameters

All Superinterfaces:

Parameters, ProtoSerializable<org.tribuo.math.protos.ParametersProto>, Serializable

All Known Implementing Classes:

FMParameters, LinearParameters

public interface FeedForwardParameters
extends Parameters

A Parameters for models which make a single prediction like logistic regressions and neural networks.

Field Summary

Fields inherited from interface org.tribuo.protos.ProtoSerializable

DESERIALIZATION_METHOD_NAME, PROVENANCE_SERIALIZER

Method Summary

Modifier and Type	Method	Description
FeedForwardParameters	copy()	Returns a copy of the parameters.
Tensor[]	gradients(com.oracle.labs.mlrg.olcut.util.Pair<Double,SGDVector> score, SGDVector features)	Generates the parameter gradients given the loss, output gradient and input features.
DenseVector	predict(SGDVector example)	Generates an un-normalized prediction by feeding the features through the parameters.

Methods inherited from interface org.tribuo.math.Parameters

get, getEmptyCopy, merge, set, update

Methods inherited from interface org.tribuo.protos.ProtoSerializable

serialize

Method Details

predict

DenseVector predict(SGDVector example)

Generates an un-normalized prediction by feeding the features through the parameters.

Parameters:

example - The features.

Returns:

The prediction.

gradients

Tensor[] gradients(com.oracle.labs.mlrg.olcut.util.Pair<Double,SGDVector> score,
 SGDVector features)

Generates the parameter gradients given the loss, output gradient and input features.

Parameters:

score - The loss and gradient.

features - The input features.

Returns:

The parameter gradient array.

copy

FeedForwardParameters copy()

Returns a copy of the parameters.

Returns:

A copy of the model parameters.