org.datavec.api.util.ndarray

Class RecordConverter

java.lang.Object

org.datavec.api.util.ndarray.RecordConverter

public class RecordConverter
extends Object

Author:

Adam Gibson

Method Summary

Modifier and Type	Method and Description
static INDArray	toArray(Collection<? extends Writable> record) Convert a record to an INDArray.
static INDArray	toArray(DataType dataType, Collection<? extends Writable> record) Convert a record to an INDArray.
static INDArray	toArray(DataType dataType, Collection<Writable> record, int size) Convert a record to an ndarray
static INDArray	toMatrix(DataType dataType, List<List<Writable>> records) Convert a set of records in to a matrix
static INDArray	toMatrix(List<List<Writable>> records) Convert a set of records in to a matrix As per toMatrix(DataType, List) but hardcoded to Float datatype
static INDArray	toMinibatchArray(@NonNull List<? extends Writable> l) Convert a record to an INDArray, for use in minibatch training.
static List<Writable>	toRecord(INDArray array) Convert an ndarray to a record
static List<Writable>	toRecord(Schema schema, List<Object> source) Convert a collection into a `List`, i.e.
static List<List<Writable>>	toRecords(DataSet dataSet) Convert a DataSet to a matrix
static List<List<Writable>>	toRecords(INDArray matrix) Convert a set of records in to a matrix
static INDArray	toTensor(List<List<List<Writable>>> records) Convert a set of records in to a matrix

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

toArray

public static INDArray toArray(DataType dataType,
                               Collection<Writable> record,
                               int size)

Convert a record to an ndarray

Parameters:

record - the record to convert

Returns:

the array

toRecords

public static List<List<Writable>> toRecords(INDArray matrix)

Convert a set of records in to a matrix

Parameters:

matrix - the records ot convert

Returns:

the matrix for the records

toTensor

public static INDArray toTensor(List<List<List<Writable>>> records)

Convert a set of records in to a matrix

Parameters:

records - the records ot convert

Returns:

the matrix for the records

toMatrix

public static INDArray toMatrix(List<List<Writable>> records)

Convert a set of records in to a matrix As per toMatrix(DataType, List) but hardcoded to Float datatype

Parameters:

records - the records ot convert

Returns:

the matrix for the records

toMatrix

public static INDArray toMatrix(DataType dataType,
                                List<List<Writable>> records)

Convert a set of records in to a matrix

Parameters:

records - the records ot convert

Returns:

the matrix for the records

toArray

public static INDArray toArray(Collection<? extends Writable> record)

Convert a record to an INDArray. May contain a mix of Writables and row vector NDArrayWritables. As per toArray(DataType, Collection) but hardcoded to Float datatype

Parameters:

record - the record to convert

Returns:

the array

toArray

public static INDArray toArray(DataType dataType,
                               Collection<? extends Writable> record)

Convert a record to an INDArray. May contain a mix of Writables and row vector NDArrayWritables.

Parameters:

record - the record to convert

Returns:

the array

toMinibatchArray

public static INDArray toMinibatchArray(@NonNull
                                        @NonNull List<? extends Writable> l)

Convert a record to an INDArray, for use in minibatch training. That is, for an input record of length N, the output array has dimension 0 of size N (i.e., suitable for minibatch training in DL4J, for example).
The input list of writables must all be the same type (i.e., all NDArrayWritables or all non-array writables such as DoubleWritable etc).
Note that for NDArrayWritables, they must have leading dimension 1, and all other dimensions must match.
For example, row vectors are valid NDArrayWritables, as are 3d (usually time series) with shape [1, x, y], or 4d (usually images) with shape [1, x, y, z] where (x,y,z) are the same for all inputs

Parameters:

l - the records to convert

Returns:

the array

See Also:

for the "single example concatenation" version of this method

toRecord

public static List<Writable> toRecord(INDArray array)

Convert an ndarray to a record

Parameters:

array - the array to convert

Returns:

the record

toRecord

public static List<Writable> toRecord(Schema schema,
                                      List<Object> source)

Convert a collection into a `List`, i.e. a record that can be used with other datavec methods. Uses a schema to decide what kind of writable to use.

Returns:

a record

toRecords

public static List<List<Writable>> toRecords(DataSet dataSet)

Convert a DataSet to a matrix

Parameters:

dataSet - the DataSet to convert

Returns:

the matrix for the records