tensor

package
v0.3.3 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Sep 9, 2024 License: BSD-3-Clause Imports: 18 Imported by: 50

README

Tensor

Tensor and related sub-packages provide a simple yet powerful framework for representing n-dimensional data of various types, providing similar functionality to the widely used numpy and pandas libraries in python, and the commercial MATLAB framework.

  • table organizes multiple Tensors as columns in a data Table, aligned by a common row dimension as the outer-most dimension of each tensor. Because the columns are tensors, each cell (value associated with a given row) can also be n-dimensional, allowing efficient representation of patterns and other high-dimensional data. Furthermore, the entire column is organized as a single contiguous slice of data, so it can be efficiently processed. The table package also has an IndexView that provides an indexed view into the rows of the table for highly efficient filtering and sorting of data.

    Data that is encoded as a slice of structs can be bidirectionally converted to / from a Table, which then provides more powerful sorting, filtering and other functionality, including the plotcore.

  • tensorcore provides core widgets for the Tensor and Table data.

  • stats implements a number of different ways of analyzing tensor and table data.

  • plot/plotcore supports interactive plotting of Table data.

History

This package was originally developed as etable as part of the emergent software framework. It always depended on the GUI framework that became Cogent Core, and having it integrated within the Core monorepo makes it easier to integrate updates, and also makes it easier to build advanced data management and visualization applications. For example, the plot/plotcore package uses the Table to support flexible and powerful plotting functionality.

Documentation

Index

Constants

View Source 
const (
	// OddRow is for oddRow arguments to Projection2D functions,
	// specifies that the odd dimension goes along the row.
	OddRow = true

	// OddColumn is for oddRow arguments to Projection2D functions,
	// specifies that the odd dimension goes along the column.
	OddColumn = false
)

Variables

This section is empty.

Functions

func BoolToFloat64 

func BoolToFloat64(bv bool) float64

func ColMajorStrides 

func ColMajorStrides(sizes []int) []int

ColMajorStrides returns strides for sizes where the first dimension is inner-most and subsequent dimensions are progressively outer

func CopyDense 

func CopyDense(to Tensor, dm *mat.Dense)

CopyDense copies a gonum mat.Dense matrix into given Tensor using standard Float64 interface

func EqualInts 

func EqualInts(a, b []int) bool

EqualInts compares two int slices and returns true if they are equal

func Float64ToBool 

func Float64ToBool(val float64) bool

func Float64ToString 

func Float64ToString(val float64) string

Float64ToString converts float64 to string value using strconv, g format

func OpenCSV added in v0.1.2 

func OpenCSV(tsr Tensor, filename core.Filename, delim rune) error

OpenCSV reads a tensor from a comma-separated-values (CSV) file (where comma = any delimiter, specified in the delim arg), using the Go standard encoding/csv reader conforming to the official CSV standard. Reads all values and assigns as many as fit.

func Projection2DCoords added in v0.1.3 

func Projection2DCoords(shp *Shape, oddRow bool, row, col int) (rowCoords, colCoords []int)

Projection2DCoords returns the corresponding full-dimensional coordinates that go into the given row, col coords for a 2D projection of the given tensor, collapsing higher dimensions down to 2D (and 1D up to 2D).

func Projection2DIndex added in v0.1.3 

func Projection2DIndex(shp *Shape, oddRow bool, row, col int) int

Projection2DIndex returns the flat 1D index for given row, col coords for a 2D projection of the given tensor shape, collapsing higher dimensions down to 2D (and 1D up to 2D). For any odd number of dimensions, the remaining outer-most dimension can either be multipliexed across the row or column, given the oddRow arg. Even multiples of inner-most dimensions are assumed to be row, then column.

func Projection2DSet added in v0.1.3 

func Projection2DSet(tsr Tensor, oddRow bool, row, col int, val float64)

Projection2DSet sets a float64 value at given row, col coords for a 2D projection of the given tensor, collapsing higher dimensions down to 2D (and 1D up to 2D). For any odd number of dimensions, the remaining outer-most dimension can either be multipliexed across the row or column, given the oddRow arg. Even multiples of inner-most dimensions are assumed to be row, then column.

func Projection2DSetString added in v0.3.3 

func Projection2DSetString(tsr Tensor, oddRow bool, row, col int, val string)

Projection2DSetString sets a string value at given row, col coords for a 2D projection of the given tensor, collapsing higher dimensions down to 2D (and 1D up to 2D). For any odd number of dimensions, the remaining outer-most dimension can either be multipliexed across the row or column, given the oddRow arg. Even multiples of inner-most dimensions are assumed to be row, then column.

func Projection2DShape added in v0.1.3 

func Projection2DShape(shp *Shape, oddRow bool) (rows, cols, rowEx, colEx int)

Projection2DShape returns the size of a 2D projection of the given tensor Shape, collapsing higher dimensions down to 2D (and 1D up to 2D). For any odd number of dimensions, the remaining outer-most dimension can either be multipliexed across the row or column, given the oddRow arg. Even multiples of inner-most dimensions are assumed to be row, then column. rowEx returns the number of "extra" (higher dimensional) rows and colEx returns the number of extra cols

func Projection2DString added in v0.3.3 

func Projection2DString(tsr Tensor, oddRow bool, row, col int) string

Projection2DString returns the string value at given row, col coords for a 2D projection of the given tensor, collapsing higher dimensions down to 2D (and 1D up to 2D). For any odd number of dimensions, the remaining outer-most dimension can either be multipliexed across the row or column, given the oddRow arg. Even multiples of inner-most dimensions are assumed to be row, then column.

func Projection2DValue added in v0.1.3 

func Projection2DValue(tsr Tensor, oddRow bool, row, col int) float64

Projection2DValue returns the float64 value at given row, col coords for a 2D projection of the given tensor, collapsing higher dimensions down to 2D (and 1D up to 2D). For any odd number of dimensions, the remaining outer-most dimension can either be multipliexed across the row or column, given the oddRow arg. Even multiples of inner-most dimensions are assumed to be row, then column.

func ReadCSV added in v0.1.2 

func ReadCSV(tsr Tensor, r io.Reader, delim rune) error

ReadCSV reads a tensor from a comma-separated-values (CSV) file (where comma = any delimiter, specified in the delim arg), using the Go standard encoding/csv reader conforming to the official CSV standard. Reads all values and assigns as many as fit.

func RowMajorStrides 

func RowMajorStrides(sizes []int) []int

RowMajorStrides returns strides for sizes where the first dimension is outer-most and subsequent dimensions are progressively inner.

func SaveCSV added in v0.1.2 

func SaveCSV(tsr Tensor, filename core.Filename, delim rune) error

SaveCSV writes a tensor to a comma-separated-values (CSV) file (where comma = any delimiter, specified in the delim arg). Outer-most dims are rows in the file, and inner-most is column -- Reading just grabs all values and doesn't care about shape.

func StringToFloat64 

func StringToFloat64(str string) float64

StringToFloat64 converts string value to float64 using strconv, returning 0 if any error

func WriteCSV added in v0.1.2 

func WriteCSV(tsr Tensor, w io.Writer, delim rune) error

WriteCSV writes a tensor to a comma-separated-values (CSV) file (where comma = any delimiter, specified in the delim arg). Outer-most dims are rows in the file, and inner-most is column -- Reading just grabs all values and doesn't care about shape.

Types

type Base 

type Base[T any] struct {
	Shp    Shape
	Values []T
	Meta   map[string]string
}

Base is an n-dim array of float64s.

func (*Base[T]) Bytes added in v0.3.3 

func (tsr *Base[T]) Bytes() []byte

func (*Base[T]) CopyMetaData 

func (tsr *Base[T]) CopyMetaData(frm Tensor)

CopyMetaData copies meta data from given source tensor

func (*Base[T]) DataType 

func (tsr *Base[T]) DataType() reflect.Kind

DataType returns the type of the data elements in the tensor. Bool is returned for the Bits tensor type.

func (*Base[T]) DimSize 

func (tsr *Base[T]) DimSize(dim int) int

DimSize returns size of given dimension

func (*Base[T]) Dims 

func (tsr *Base[T]) Dims() (r, c int)

Dims is the gonum/mat.Matrix interface method for returning the dimensionality of the 2D Matrix. Assumes Row-major ordering and logs an error if NumDims < 2.

func (*Base[T]) Label 

func (tsr *Base[T]) Label() string

Label satisfies the core.Labeler interface for a summary description of the tensor

func (*Base[T]) Len 

func (tsr *Base[T]) Len() int

Len returns the number of elements in the tensor (product of shape dimensions).

func (*Base[T]) MetaData 

func (tsr *Base[T]) MetaData(key string) (string, bool)

MetaData retrieves value of given key, bool = false if not set

func (*Base[T]) MetaDataMap 

func (tsr *Base[T]) MetaDataMap() map[string]string

MetaDataMap returns the underlying map used for meta data

func (*Base[T]) NumDims 

func (tsr *Base[T]) NumDims() int

NumDims returns the total number of dimensions.

func (*Base[T]) RowCellSize 

func (tsr *Base[T]) RowCellSize() (rows, cells int)

RowCellSize returns the size of the outer-most Row shape dimension, and the size of all the remaining inner dimensions (the "cell" size). Used for Tensors that are columns in a data table.

func (*Base[T]) Set 

func (tsr *Base[T]) Set(i []int, val T)

func (*Base[T]) Set1D 

func (tsr *Base[T]) Set1D(i int, val T)

func (*Base[T]) SetMetaData 

func (tsr *Base[T]) SetMetaData(key, val string)

SetMetaData sets a key=value meta data (stored as a map[string]string). For TensorGrid display: top-zero=+/-, odd-row=+/-, image=+/-, min, max set fixed min / max values, background=color

func (*Base[T]) SetNumRows 

func (tsr *Base[T]) SetNumRows(rows int)

SetNumRows sets the number of rows (outer-most dimension) in a RowMajor organized tensor.

func (*Base[T]) SetShape 

func (tsr *Base[T]) SetShape(sizes []int, names ...string)

SetShape sets the shape params, resizing backing storage appropriately

func (*Base[T]) Shape 

func (tsr *Base[T]) Shape() *Shape

Shape returns a pointer to the shape that fully parametrizes the tensor shape

func (*Base[T]) Sizeof added in v0.3.3 

func (tsr *Base[T]) Sizeof() int64

func (*Base[T]) String1D 

func (tsr *Base[T]) String1D(off int) string

func (*Base[T]) StringRowCell 

func (tsr *Base[T]) StringRowCell(row, cell int) string

func (*Base[T]) StringValue 

func (tsr *Base[T]) StringValue(i []int) string

func (*Base[T]) Symmetric 

func (tsr *Base[T]) Symmetric() (r int)

Symmetric is the gonum/mat.Matrix interface method for returning the dimensionality of a symmetric 2D Matrix.

func (*Base[T]) SymmetricDim 

func (tsr *Base[T]) SymmetricDim() int

SymmetricDim returns the number of rows/columns in the matrix.

func (*Base[T]) Value 

func (tsr *Base[T]) Value(i []int) T

func (*Base[T]) Value1D 

func (tsr *Base[T]) Value1D(i int) T

type Bits 

type Bits struct {
	Shp    Shape
	Values bitslice.Slice
	Meta   map[string]string
}

Bits is a tensor of bits backed by a bitslice.Slice for efficient storage of binary data

func NewBits 

func NewBits(sizes []int, names ...string) *Bits

NewBits returns a new n-dimensional tensor of bit values with the given sizes per dimension (shape), and optional dimension names.

func NewBitsShape 

func NewBitsShape(shape *Shape) *Bits

NewBitsShape returns a new n-dimensional tensor of bit values using given shape.

func (*Bits) At 

func (tsr *Bits) At(i, j int) float64

At is the gonum/mat.Matrix interface method for returning 2D matrix element at given row, column index. Not supported for Bits -- do not call!

func (*Bits) Bytes added in v0.3.3 

func (tsr *Bits) Bytes() []byte

func (*Bits) Clone 

func (tsr *Bits) Clone() Tensor

Clone clones this tensor, creating a duplicate copy of itself with its own separate memory representation of all the values, and returns that as a Tensor (which can be converted into the known type as needed).

func (*Bits) CopyCellsFrom 

func (tsr *Bits) CopyCellsFrom(frm Tensor, to, start, n int)

CopyCellsFrom copies given range of values from other tensor into this tensor, using flat 1D indexes: to = starting index in this Tensor to start copying into, start = starting index on from Tensor to start copying from, and n = number of values to copy. Uses an optimized implementation if the other tensor is of the same type, and otherwise it goes through appropriate standard type.

func (*Bits) CopyFrom 

func (tsr *Bits) CopyFrom(frm Tensor)

CopyFrom copies all avail values from other tensor into this tensor, with an optimized implementation if the other tensor is of the same type, and otherwise it goes through appropriate standard type.

func (*Bits) CopyMetaData 

func (tsr *Bits) CopyMetaData(frm Tensor)

CopyMetaData copies meta data from given source tensor

func (*Bits) CopyShapeFrom 

func (tsr *Bits) CopyShapeFrom(frm Tensor)

CopyShapeFrom copies just the shape from given source tensor calling SetShape with the shape params from source (see for more docs).

func (*Bits) DataType 

func (tsr *Bits) DataType() reflect.Kind

DataType returns the type of the data elements in the tensor. Bool is returned for the Bits tensor type.

func (*Bits) DimSize 

func (tsr *Bits) DimSize(dim int) int

DimSize returns size of given dimension

func (*Bits) Dims 

func (tsr *Bits) Dims() (r, c int)

Dims is the gonum/mat.Matrix interface method for returning the dimensionality of the 2D Matrix. Not supported for Bits -- do not call!

func (*Bits) Float 

func (tsr *Bits) Float(i []int) float64

func (*Bits) Float1D 

func (tsr *Bits) Float1D(off int) float64

func (*Bits) FloatRowCell 

func (tsr *Bits) FloatRowCell(row, cell int) float64

func (*Bits) Floats 

func (tsr *Bits) Floats(flt *[]float64)

func (*Bits) IsString 

func (tsr *Bits) IsString() bool

func (*Bits) Label 

func (tsr *Bits) Label() string

Label satisfies the core.Labeler interface for a summary description of the tensor

func (*Bits) Len 

func (tsr *Bits) Len() int

Len returns the number of elements in the tensor (product of shape dimensions).

func (*Bits) MetaData 

func (tsr *Bits) MetaData(key string) (string, bool)

MetaData retrieves value of given key, bool = false if not set

func (*Bits) MetaDataMap 

func (tsr *Bits) MetaDataMap() map[string]string

MetaDataMap returns the underlying map used for meta data

func (*Bits) NumDims 

func (tsr *Bits) NumDims() int

NumDims returns the total number of dimensions.

func (*Bits) Range 

func (tsr *Bits) Range() (min, max float64, minIndex, maxIndex int)

Range is not applicable to Bits tensor

func (*Bits) RowCellSize 

func (tsr *Bits) RowCellSize() (rows, cells int)

RowCellSize returns the size of the outer-most Row shape dimension, and the size of all the remaining inner dimensions (the "cell" size). Used for Tensors that are columns in a data table.

func (*Bits) Set 

func (tsr *Bits) Set(i []int, val bool)

func (*Bits) Set1D 

func (tsr *Bits) Set1D(i int, val bool)

func (*Bits) SetFloat 

func (tsr *Bits) SetFloat(i []int, val float64)

func (*Bits) SetFloat1D 

func (tsr *Bits) SetFloat1D(off int, val float64)

func (*Bits) SetFloatRowCell 

func (tsr *Bits) SetFloatRowCell(row, cell int, val float64)

func (*Bits) SetFloats 

func (tsr *Bits) SetFloats(vals []float64)

SetFloats sets tensor values from a []float64 slice (copies values).

func (*Bits) SetMetaData 

func (tsr *Bits) SetMetaData(key, val string)

SetMetaData sets a key=value meta data (stored as a map[string]string). For TensorGrid display: top-zero=+/-, odd-row=+/-, image=+/-, min, max set fixed min / max values, background=color

func (*Bits) SetNumRows 

func (tsr *Bits) SetNumRows(rows int)

SetNumRows sets the number of rows (outer-most dimension) in a RowMajor organized tensor.

func (*Bits) SetShape 

func (tsr *Bits) SetShape(sizes []int, names ...string)

SetShape sets the shape params, resizing backing storage appropriately

func (*Bits) SetString 

func (tsr *Bits) SetString(i []int, val string)

func (*Bits) SetString1D 

func (tsr *Bits) SetString1D(off int, val string)

func (*Bits) SetStringRowCell 

func (tsr *Bits) SetStringRowCell(row, cell int, val string)

func (*Bits) SetZeros 

func (tsr *Bits) SetZeros()

SetZeros is simple convenience function initialize all values to 0

func (*Bits) Shape 

func (tsr *Bits) Shape() *Shape

Shape returns a pointer to the shape that fully parametrizes the tensor shape

func (*Bits) Sizeof added in v0.3.3 

func (tsr *Bits) Sizeof() int64

func (*Bits) String 

func (tsr *Bits) String() string

String satisfies the fmt.Stringer interface for string of tensor data

func (*Bits) String1D 

func (tsr *Bits) String1D(off int) string

func (*Bits) StringRowCell 

func (tsr *Bits) StringRowCell(row, cell int) string

func (*Bits) StringValue 

func (tsr *Bits) StringValue(i []int) string

func (*Bits) SubSpace 

func (tsr *Bits) SubSpace(offs []int) Tensor

SubSpace is not possible with Bits

func (*Bits) T 

func (tsr *Bits) T() mat.Matrix

T is the gonum/mat.Matrix transpose method. Not supported for Bits -- do not call!

func (*Bits) Value 

func (tsr *Bits) Value(i []int) bool

Value returns value at given tensor index

func (*Bits) Value1D 

func (tsr *Bits) Value1D(i int) bool

Value1D returns value at given tensor 1D (flat) index

type Byte 

type Byte = Number[byte]

Byte is an alias for Number[byte].

func NewByte added in v0.1.2 

func NewByte(sizes []int, names ...string) *Byte

NewByte returns a new Byte tensor with the given sizes per dimension (shape), and optional dimension names.

type Float32 

type Float32 = Number[float32]

Float32 is an alias for Number[float32].

func NewFloat32 added in v0.1.2 

func NewFloat32(sizes []int, names ...string) *Float32

NewFloat32 returns a new Float32 tensor with the given sizes per dimension (shape), and optional dimension names.

type Float64 

type Float64 = Number[float64]

Float64 is an alias for Number[float64].

func NewFloat64 added in v0.1.2 

func NewFloat64(sizes []int, names ...string) *Float64

NewFloat64 returns a new Float64 tensor with the given sizes per dimension (shape), and optional dimension names.

type Int 

type Int = Number[int]

Int is an alias for Number[int].

func NewInt added in v0.1.2 

func NewInt(sizes []int, names ...string) *Int

NewInt returns a new Int tensor with the given sizes per dimension (shape), and optional dimension names.

type Int32 

type Int32 = Number[int32]

Int32 is an alias for Number[int32].

func NewInt32 added in v0.1.2 

func NewInt32(sizes []int, names ...string) *Int32

NewInt32 returns a new Int32 tensor with the given sizes per dimension (shape), and optional dimension names.

type Number 

type Number[T num.Number] struct {
	Base[T]
}

Number is a tensor of numerical values

func NewNumber 

func NewNumber[T num.Number](sizes []int, names ...string) *Number[T]

NewNumber returns a new n-dimensional tensor of numerical values with the given sizes per dimension (shape), and optional dimension names.

func NewNumberShape 

func NewNumberShape[T num.Number](shape *Shape) *Number[T]

NewNumberShape returns a new n-dimensional tensor of numerical values using given shape.

func (*Number[T]) AddScalar 

func (tsr *Number[T]) AddScalar(i []int, val float64) float64

func (*Number[T]) At 

func (tsr *Number[T]) At(i, j int) float64

At is the gonum/mat.Matrix interface method for returning 2D matrix element at given row, column index. Assumes Row-major ordering and logs an error if NumDims < 2.

func (*Number[T]) Clone 

func (tsr *Number[T]) Clone() Tensor

Clone clones this tensor, creating a duplicate copy of itself with its own separate memory representation of all the values, and returns that as a Tensor (which can be converted into the known type as needed).

func (*Number[T]) CopyCellsFrom 

func (tsr *Number[T]) CopyCellsFrom(frm Tensor, to, start, n int)

CopyCellsFrom copies given range of values from other tensor into this tensor, using flat 1D indexes: to = starting index in this Tensor to start copying into, start = starting index on from Tensor to start copying from, and n = number of values to copy. Uses an optimized implementation if the other tensor is of the same type, and otherwise it goes through appropriate standard type.

func (*Number[T]) CopyFrom 

func (tsr *Number[T]) CopyFrom(frm Tensor)

CopyFrom copies all avail values from other tensor into this tensor, with an optimized implementation if the other tensor is of the same type, and otherwise it goes through appropriate standard type.

func (*Number[T]) CopyShapeFrom 

func (tsr *Number[T]) CopyShapeFrom(frm Tensor)

CopyShapeFrom copies just the shape from given source tensor calling SetShape with the shape params from source (see for more docs).

func (*Number[T]) Float 

func (tsr *Number[T]) Float(i []int) float64

func (*Number[T]) Float1D 

func (tsr *Number[T]) Float1D(i int) float64

func (*Number[T]) FloatRowCell 

func (tsr *Number[T]) FloatRowCell(row, cell int) float64

func (*Number[T]) Floats 

func (tsr *Number[T]) Floats(flt *[]float64)

Floats sets []float64 slice of all elements in the tensor (length is ensured to be sufficient). This can be used for all of the gonum/floats methods for basic math, gonum/stats, etc.

func (*Number[T]) IsString 

func (tsr *Number[T]) IsString() bool

func (*Number[T]) MulScalar 

func (tsr *Number[T]) MulScalar(i []int, val float64) float64

func (*Number[T]) Range 

func (tsr *Number[T]) Range() (min, max float64, minIndex, maxIndex int)

Range returns the min, max (and associated indexes, -1 = no values) for the tensor. This is needed for display and is thus in the core api in optimized form Other math operations can be done using gonum/floats package.

func (*Number[T]) SetFloat 

func (tsr *Number[T]) SetFloat(i []int, val float64)

func (*Number[T]) SetFloat1D 

func (tsr *Number[T]) SetFloat1D(i int, val float64)

func (*Number[T]) SetFloatRowCell 

func (tsr *Number[T]) SetFloatRowCell(row, cell int, val float64)

func (*Number[T]) SetFloats 

func (tsr *Number[T]) SetFloats(flt []float64)

SetFloats sets tensor values from a []float64 slice (copies values).

func (*Number[T]) SetString 

func (tsr *Number[T]) SetString(i []int, val string)

func (Number[T]) SetString1D 

func (tsr Number[T]) SetString1D(off int, val string)

func (*Number[T]) SetStringRowCell 

func (tsr *Number[T]) SetStringRowCell(row, cell int, val string)

func (*Number[T]) SetZeros 

func (tsr *Number[T]) SetZeros()

SetZeros is simple convenience function initialize all values to 0

func (*Number[T]) String 

func (tsr *Number[T]) String() string

String satisfies the fmt.Stringer interface for string of tensor data

func (*Number[T]) SubSpace 

func (tsr *Number[T]) SubSpace(offs []int) Tensor

SubSpace returns a new tensor with innermost subspace at given offset(s) in outermost dimension(s) (len(offs) < NumDims). The new tensor points to the values of the this tensor (i.e., modifications will affect both), as its Values slice is a view onto the original (which is why only inner-most contiguous supsaces are supported). Use Clone() method to separate the two.

func (*Number[T]) T 

func (tsr *Number[T]) T() mat.Matrix

T is the gonum/mat.Matrix transpose method. It performs an implicit transpose by returning the receiver inside a Transpose.

type Shape 

type Shape struct {

	// size per dimension
	Sizes []int

	// offsets for each dimension
	Strides []int `display:"-"`

	// names of each dimension
	Names []string `display:"-"`
}

Shape manages a tensor's shape information, including strides and dimension names and can compute the flat index into an underlying 1D data storage array based on an n-dimensional index (and vice-versa). Per C / Go / Python conventions, indexes are Row-Major, ordered from outer to inner left-to-right, so the inner-most is right-most.

func AddShapes added in v0.1.2 

func AddShapes(shape1, shape2 *Shape) *Shape

AddShapes returns a new shape by adding two shapes one after the other.

func NewShape 

func NewShape(sizes []int, names ...string) *Shape

NewShape returns a new shape with given sizes and optional dimension names. RowMajor ordering is used by default.

func (*Shape) CopyShape 

func (sh *Shape) CopyShape(cp *Shape)

CopyShape copies the shape parameters from another Shape struct. copies the data so it is not accidentally subject to updates.

func (*Shape) DimByName 

func (sh *Shape) DimByName(name string) int

DimByName returns the index of the given dimension name. returns -1 if not found.

func (*Shape) DimName 

func (sh *Shape) DimName(i int) string

DimName returns the name of given dimension.

func (*Shape) DimSize added in v0.1.2 

func (sh *Shape) DimSize(i int) int

DimSize returns the size of given dimension.

func (*Shape) DimSizeByName added in v0.1.2 

func (sh *Shape) DimSizeByName(name string) int

DimSizeByName returns the size of given dimension, specified by name. will crash if name not found.

func (*Shape) Index 

func (sh *Shape) Index(offset int) []int

Index returns the n-dimensional index from a "flat" 1D array index.

func (*Shape) IndexIsValid 

func (sh *Shape) IndexIsValid(idx []int) bool

IndexIsValid() returns true if given index is valid (within ranges for all dimensions)

func (*Shape) IsEqual 

func (sh *Shape) IsEqual(oth *Shape) bool

IsEqual returns true if this shape is same as other (does not compare names)

func (*Shape) Len 

func (sh *Shape) Len() int

Len returns the total length of elements in the tensor (i.e., the product of the shape sizes)

func (*Shape) NumDims 

func (sh *Shape) NumDims() int

NumDims returns the total number of dimensions.

func (*Shape) Offset 

func (sh *Shape) Offset(index []int) int

Offset returns the "flat" 1D array index into an element at the given n-dimensional index. No checking is done on the length or size of the index values relative to the shape of the tensor.

func (*Shape) RowCellSize 

func (sh *Shape) RowCellSize() (rows, cells int)

RowCellSize returns the size of the outer-most Row shape dimension, and the size of all the remaining inner dimensions (the "cell" size). Used for Tensors that are columns in a data table.

func (*Shape) SetShape 

func (sh *Shape) SetShape(sizes []int, names ...string)

SetShape sets the shape size and optional names RowMajor ordering is used by default.

func (*Shape) String 

func (sh *Shape) String() string

String satisfies the fmt.Stringer interface

type String 

type String struct {
	Base[string]
}

String is a tensor of string values

func NewString 

func NewString(sizes []int, names ...string) *String

NewString returns a new n-dimensional tensor of string values with the given sizes per dimension (shape), and optional dimension names.

func NewStringShape 

func NewStringShape(shape *Shape) *String

NewStringShape returns a new n-dimensional tensor of string values using given shape.

func (*String) AddScalar 

func (tsr *String) AddScalar(i []int, val float64) float64

func (*String) At 

func (tsr *String) At(i, j int) float64

At is the gonum/mat.Matrix interface method for returning 2D matrix element at given row, column index. Assumes Row-major ordering and logs an error if NumDims < 2.

func (*String) Clone 

func (tsr *String) Clone() Tensor

Clone clones this tensor, creating a duplicate copy of itself with its own separate memory representation of all the values, and returns that as a Tensor (which can be converted into the known type as needed).

func (*String) CopyCellsFrom 

func (tsr *String) CopyCellsFrom(frm Tensor, to, start, n int)

CopyCellsFrom copies given range of values from other tensor into this tensor, using flat 1D indexes: to = starting index in this Tensor to start copying into, start = starting index on from Tensor to start copying from, and n = number of values to copy. Uses an optimized implementation if the other tensor is of the same type, and otherwise it goes through appropriate standard type.

func (*String) CopyFrom 

func (tsr *String) CopyFrom(frm Tensor)

CopyFrom copies all avail values from other tensor into this tensor, with an optimized implementation if the other tensor is of the same type, and otherwise it goes through appropriate standard type.

func (*String) CopyShapeFrom 

func (tsr *String) CopyShapeFrom(frm Tensor)

CopyShapeFrom copies just the shape from given source tensor calling SetShape with the shape params from source (see for more docs).

func (*String) Float 

func (tsr *String) Float(i []int) float64

func (*String) Float1D 

func (tsr *String) Float1D(off int) float64

func (*String) FloatRowCell 

func (tsr *String) FloatRowCell(row, cell int) float64

func (*String) Floats 

func (tsr *String) Floats(flt *[]float64)

Floats sets []float64 slice of all elements in the tensor (length is ensured to be sufficient). This can be used for all of the gonum/floats methods for basic math, gonum/stats, etc.

func (*String) IsString 

func (tsr *String) IsString() bool

func (*String) MulScalar 

func (tsr *String) MulScalar(i []int, val float64) float64

func (*String) Range 

func (tsr *String) Range() (min, max float64, minIndex, maxIndex int)

Range returns the min, max (and associated indexes, -1 = no values) for the tensor. This is needed for display and is thus in the core api in optimized form Other math operations can be done using gonum/floats package.

func (*String) SetFloat 

func (tsr *String) SetFloat(i []int, val float64)

func (*String) SetFloat1D 

func (tsr *String) SetFloat1D(off int, val float64)

func (*String) SetFloatRowCell 

func (tsr *String) SetFloatRowCell(row, cell int, val float64)

func (*String) SetFloats 

func (tsr *String) SetFloats(flt []float64)

SetFloats sets tensor values from a []float64 slice (copies values).

func (*String) SetString 

func (tsr *String) SetString(i []int, val string)

func (String) SetString1D 

func (tsr String) SetString1D(off int, val string)

func (*String) SetStringRowCell 

func (tsr *String) SetStringRowCell(row, cell int, val string)

func (*String) SetZeros 

func (tsr *String) SetZeros()

SetZeros is simple convenience function initialize all values to 0

func (*String) String 

func (tsr *String) String() string

String satisfies the fmt.Stringer interface for string of tensor data

func (*String) SubSpace 

func (tsr *String) SubSpace(offs []int) Tensor

SubSpace returns a new tensor with innermost subspace at given offset(s) in outermost dimension(s) (len(offs) < NumDims). The new tensor points to the values of the this tensor (i.e., modifications will affect both), as its Values slice is a view onto the original (which is why only inner-most contiguous supsaces are supported). Use Clone() method to separate the two.

func (*String) T 

func (tsr *String) T() mat.Matrix

T is the gonum/mat.Matrix transpose method. It performs an implicit transpose by returning the receiver inside a Transpose.

type Tensor 

type Tensor interface {
	fmt.Stringer
	mat.Matrix

	// Shape returns a pointer to the shape that fully parametrizes the tensor shape
	Shape() *Shape

	// Len returns the number of elements in the tensor (product of shape dimensions).
	Len() int

	// NumDims returns the total number of dimensions.
	NumDims() int

	// DimSize returns size of given dimension
	DimSize(dim int) int

	// RowCellSize returns the size of the outer-most Row shape dimension,
	// and the size of all the remaining inner dimensions (the "cell" size).
	// Used for Tensors that are columns in a data table.
	RowCellSize() (rows, cells int)

	// DataType returns the type of the data elements in the tensor.
	// Bool is returned for the Bits tensor type.
	DataType() reflect.Kind

	// Sizeof returns the number of bytes contained in the Values of this tensor.
	// for String types, this is just the string pointers.
	Sizeof() int64

	// Bytes returns the underlying byte representation of the tensor values.
	// This is the actual underlying data, so make a copy if it can be
	// unintentionally modified or retained more than for immediate use.
	Bytes() []byte

	// returns true if the data type is a String. otherwise is numeric.
	IsString() bool

	// Float returns the value of given index as a float64.
	Float(i []int) float64

	// SetFloat sets the value of given index as a float64
	SetFloat(i []int, val float64)

	// StringValue returns the value of given index as a string
	StringValue(i []int) string

	// SetString sets the value of given index as a string
	SetString(i []int, val string)

	// Float1D returns the value of given 1-dimensional index (0-Len()-1) as a float64
	Float1D(i int) float64

	// SetFloat1D sets the value of given 1-dimensional index (0-Len()-1) as a float64
	SetFloat1D(i int, val float64)

	// FloatRowCell returns the value at given row and cell, where row is outer-most dim,
	// and cell is 1D index into remaining inner dims. For Table columns.
	FloatRowCell(row, cell int) float64

	// SetFloatRowCell sets the value at given row and cell, where row is outer-most dim,
	// and cell is 1D index into remaining inner dims. For Table columns.
	SetFloatRowCell(row, cell int, val float64)

	// Floats sets []float64 slice of all elements in the tensor
	// (length is ensured to be sufficient).
	// This can be used for all of the gonum/floats methods
	// for basic math, gonum/stats, etc.
	Floats(flt *[]float64)

	// SetFloats sets tensor values from a []float64 slice (copies values).
	SetFloats(vals []float64)

	// String1D returns the value of given 1-dimensional index (0-Len()-1) as a string
	String1D(i int) string

	// SetString1D sets the value of given 1-dimensional index (0-Len()-1) as a string
	SetString1D(i int, val string)

	// StringRowCell returns the value at given row and cell, where row is outer-most dim,
	// and cell is 1D index into remaining inner dims. For Table columns
	StringRowCell(row, cell int) string

	// SetStringRowCell sets the value at given row and cell, where row is outer-most dim,
	// and cell is 1D index into remaining inner dims. For Table columns
	SetStringRowCell(row, cell int, val string)

	// SubSpace returns a new tensor with innermost subspace at given
	// offset(s) in outermost dimension(s) (len(offs) < NumDims).
	// The new tensor points to the values of the this tensor (i.e., modifications
	// will affect both), as its Values slice is a view onto the original (which
	// is why only inner-most contiguous supsaces are supported).
	// Use Clone() method to separate the two.
	SubSpace(offs []int) Tensor

	// Range returns the min, max (and associated indexes, -1 = no values) for the tensor.
	// This is needed for display and is thus in the core api in optimized form
	// Other math operations can be done using gonum/floats package.
	Range() (min, max float64, minIndex, maxIndex int)

	// SetZeros is simple convenience function initialize all values to 0
	SetZeros()

	// Clone clones this tensor, creating a duplicate copy of itself with its
	// own separate memory representation of all the values, and returns
	// that as a Tensor (which can be converted into the known type as needed).
	Clone() Tensor

	// CopyFrom copies all avail values from other tensor into this tensor, with an
	// optimized implementation if the other tensor is of the same type, and
	// otherwise it goes through appropriate standard type.
	CopyFrom(from Tensor)

	// CopyShapeFrom copies just the shape from given source tensor
	// calling SetShape with the shape params from source (see for more docs).
	CopyShapeFrom(from Tensor)

	// CopyCellsFrom copies given range of values from other tensor into this tensor,
	// using flat 1D indexes: to = starting index in this Tensor to start copying into,
	// start = starting index on from Tensor to start copying from, and n = number of
	// values to copy.  Uses an optimized implementation if the other tensor is
	// of the same type, and otherwise it goes through appropriate standard type.
	CopyCellsFrom(from Tensor, to, start, n int)

	// SetShape sets the sizes parameters of the tensor, and resizes backing storage appropriately.
	// existing names will be preserved if not presented.
	SetShape(sizes []int, names ...string)

	// SetNumRows sets the number of rows (outer-most dimension).
	SetNumRows(rows int)

	// SetMetaData sets a key=value meta data (stored as a map[string]string).
	// For TensorGrid display: top-zero=+/-, odd-row=+/-, image=+/-,
	// min, max set fixed min / max values, background=color
	SetMetaData(key, val string)

	// MetaData retrieves value of given key, bool = false if not set
	MetaData(key string) (string, bool)

	// MetaDataMap returns the underlying map used for meta data
	MetaDataMap() map[string]string

	// CopyMetaData copies meta data from given source tensor
	CopyMetaData(from Tensor)
}

Tensor is the interface for n-dimensional tensors. Per C / Go / Python conventions, indexes are Row-Major, ordered from outer to inner left-to-right, so the inner-most is right-most. It is implemented by the Base and Number generic types specialized by different concrete types: float64, float32, int, int32, byte, string, bits (bools). For float32 and float64 values, use NaN to indicate missing values. All of the data analysis and plot packages skip NaNs.

func New 

func New[T string | bool | float32 | float64 | int | int32 | byte](sizes []int, names ...string) Tensor

New returns a new n-dimensional tensor of given value type with the given sizes per dimension (shape), and optional dimension names.

func NewOfType 

func NewOfType(typ reflect.Kind, sizes []int, names ...string) Tensor

NewOfType returns a new n-dimensional tensor of given reflect.Kind type with the given sizes per dimension (shape), and optional dimension names. Supported types are string, bool (for Bits), float32, float64, int, int32, and byte.

Source Files

View all Source files

Directories

Path Synopsis
Package bitslice implements a simple slice-of-bits using a []byte slice for storage, which is used for efficient storage of boolean data, such as projection connectivity patterns.
 Package bitslice implements a simple slice-of-bits using a []byte slice for storage, which is used for efficient storage of boolean data, such as projection connectivity patterns.
cmd
tablecat
ttail
examples
datafs-sim
dataproc
grids
Package agg provides aggregation functions operating on IndexView indexed views of table.Table data, along with standard AggFunc functions that can be used at any level of aggregation from tensor on up.
 Package agg provides aggregation functions operating on IndexView indexed views of table.Table data, along with standard AggFunc functions that can be used at any level of aggregation from tensor on up.
clust
convolve
glm
histogram
metric
Package metric provides various similarity / distance metrics for comparing floating-point vectors.
 Package metric provides various similarity / distance metrics for comparing floating-point vectors.
norm
Package norm provides normalization and norm metric computations e.g., L2 = sqrt of sum of squares of a vector.
 Package norm provides normalization and norm metric computations e.g., L2 = sqrt of sum of squares of a vector.
pca
Package pca performs principal component's analysis and associated covariance matrix computations, operating on table.Table or tensor.Tensor data.
 Package pca performs principal component's analysis and associated covariance matrix computations, operating on table.Table or tensor.Tensor data.
simat
Package simat provides similarity / distance matrix functions that create a SimMat matrix from Tensor or Table data.
 Package simat provides similarity / distance matrix functions that create a SimMat matrix from Tensor or Table data.
split
Package split provides GroupBy, Agg, Permute and other functions that create and populate Splits of table.Table data.
 Package split provides GroupBy, Agg, Permute and other functions that create and populate Splits of table.Table data.
Package tensorcore provides GUI Views of table Table and Tensor structures using the Cogent Core views framework.
 Package tensorcore provides GUI Views of table Table and Tensor structures using the Cogent Core views framework.
Package tensormpi has methods to support use of MPI with tensor and table data structures.
 Package tensormpi has methods to support use of MPI with tensor and table data structures.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.