Class Tensor

Tensor class

Constructors

constructor

Accessors

dimension length sizes value

Methods

_to_index _to_position [iterator] at broadcastOperate concat copy dot equals fill flip forEach index map operateAt reduce repeat reshape resize select set shuffle slice toArray toMatrix toScaler toString transpose fromArray ones randn random zeros

Constructors

constructor

  • new Tensor(size: number[], value?: number | number[]): Tensor

  • Parameters

    • size: number[]

      Sizes for each dimension

    • Optionalvalue: number | number[]

      Initial values

    Returns Tensor

Accessors

dimension

  • get dimension(): number

  • Dimension of the tensor.

    Returns number

length

  • get length(): number

  • Number of all elements in the tensor.

    Returns number

sizes

  • get sizes(): number[]

  • Sizes of the tensor.

    Returns number[]

value

  • get value(): number[]

  • Elements in the tensor.

    Returns number[]

Methods

_to_index

  • _to_index(p: any): any[]

  • Parameters

    • p: any

    Returns any[]

_to_position

  • _to_position(...i: any[]): number

  • Parameters

    • ...i: any[]

    Returns number

[iterator]

  • "[iterator]"(): Generator<void, undefined>

  • Iterate over the elements.

    Returns Generator<void, undefined>

    Yields

at

  • at(...i: number[]): number

  • Returns value at the index position.

    Parameters

    • ...i: number[]

      Index values

    Returns number

    The value

  • at(i: number[]): number

  • Returns value at the index position.

    Parameters

    • i: number[]

      Index values

    Returns number

    The value

broadcastOperate

  • broadcastOperate(
        o: number | Tensor | Matrix,
        fn: (arg0: number, arg1: number) => number,
    ): void

  • Apply function for all elements with broadcasting.

    Parameters

    • o: number | Tensor | Matrix

      Applied value

    • fn: (arg0: number, arg1: number) => number

      Applied function

    Returns void

concat

  • concat(t: Tensor, axis?: number): void

  • Concatenate this and t.

    Parameters

    • t: Tensor

      Concatenate tensor

    • Optionalaxis: number

      Axis to be concatenated

    Returns void

copy

  • copy(): Tensor

  • Returns a copy of this tensor.

    Returns Tensor

    Copied tensor

dot

  • dot(o: Matrix): Tensor

  • Returns a tensor product value.

    Parameters

    Returns Tensor

    Producted tensor

equals

  • equals(other: any): boolean

  • Returns this tensor is equals to the others.

    Parameters

    • other: any

      Check tensor

    Returns boolean

    true if equal

fill

  • fill(value: number): void

  • Fill in all the elements with the value.

    Parameters

    • value: number

      Filled value

    Returns void

flip

  • flip(axis?: number): void

  • Flip values along the axis.

    Parameters

    • Optionalaxis: number

      Axis to be flipped

    Returns void

forEach

  • forEach(cb: (arg0: number, arg1: number[], arg2: Tensor) => void): void

  • Iterate over all the elements.

    Parameters

    • cb: (arg0: number, arg1: number[], arg2: Tensor) => void

      Callback function

    Returns void

index

  • index(...i: number[]): Tensor

  • Returns tensor at the index position.

    Parameters

    • ...i: number[]

      Index values

    Returns Tensor

    Sub tensor

  • index(i: number[]): Tensor

  • Returns tensor at the index position.

    Parameters

    • i: number[]

      Index values

    Returns Tensor

    Sub tensor

map

  • map(cb: (arg0: number, arg1: number[], arg2: Tensor) => number): void

  • Iterate over all the elements and replace the value.

    Parameters

    • cb: (arg0: number, arg1: number[], arg2: Tensor) => number

      Mapping function

    Returns void

operateAt

  • operateAt(i: number | number[], fn?: (arg0: number) => number): number

  • Apply function to the position.

    Parameters

    • i: number | number[]

      Index values

    • Optionalfn: (arg0: number) => number

      Applied function

    Returns number

    Old value

reduce

  • reduce(
        cb: (arg0: number, arg1: number, arg2: number[], arg3: Tensor) => number,
        init?: any,
    ): number

  • Returns a tensor reduced along all element with the callback function.

    Parameters

    • cb: (arg0: number, arg1: number, arg2: number[], arg3: Tensor) => number

      Reducing function

    • Optionalinit: any

      Initial value

    Returns number

    Reduced tensor or value

  • reduce(
        cb: (arg0: number, arg1: number, arg2: number[], arg3: Tensor) => number,
        init: any,
        axis: number | number[],
        keepdims?: boolean,
    ): number | Tensor

  • Returns a tensor reduced along the axis with the callback function.

    Parameters

    • cb: (arg0: number, arg1: number, arg2: number[], arg3: Tensor) => number

      Reducing function

    • init: any

      Initial value

    • axis: number | number[]

      Axis to be reduced. If negative, reduce along all elements.

    • Optionalkeepdims: boolean

      Keep dimensions or not.

    Returns number | Tensor

    Reduced tensor or value

repeat

  • repeat(n: number, axis?: number): any

  • Repeat the elements n times along the axis this.

    Parameters

    • n: number

      Repeated count

    • Optionalaxis: number

      Axis to be repeated

    Returns any

  • repeat(n: number[]): any

  • Repeat the elements n times along the axis this.

    Parameters

    • n: number[]

      Repeated counts of each axis

    Returns any

reshape

  • reshape(...sizes: number[]): any

  • Reshape this as the sizes.

    Parameters

    • ...sizes: number[]

      New sizes for each dimension

    Returns any

  • reshape(sizes: number[]): any

  • Reshape this as the sizes.

    Parameters

    • sizes: number[]

      New sizes for each dimension

    Returns any

resize

  • resize(sizes: number[], init?: number): void

  • Resize this tensor.

    Parameters

    • sizes: number[]

      New sizes

    • Optionalinit: number

      Value of the extended region

    Returns void

select

  • select(idx: number | number[], axis?: number): Tensor

  • Returns the sub-tensor corresponding to position i in the first dimension of this.

    Parameters

    • idx: number | number[]

      Select index value(s)

    • Optionalaxis: number

      Axis

    Returns Tensor

    Selected tensor

set

  • set(i: number | number[], value: number): void

  • Set the value at the specific position.

    Parameters

    • i: number | number[]

      Index values

    • value: number

      Set value

    Returns void

shuffle

  • shuffle(axis?: number): void

  • Shuffle along the axis.

    Parameters

    • Optionalaxis: number

      Axis

    Returns void

slice

  • slice(from: number, to: number, axis?: number): Tensor

  • Returns a tensor sliced by first dimension.

    Parameters

    • from: number

      Start index

    • to: number

      End index

    • Optionalaxis: number

      Axis

    Returns Tensor

    Sliced tensor

toArray

  • toArray(): number[]

  • Returns a nested array represented this tensor.

    Returns number[]

    Nested array

toMatrix

  • toMatrix(): Matrix

  • Returns a Matrix if the dimension of this tensor is 2.

    Returns Matrix

    Matrix

    Throws

    If the dimension of this tensor is not 2.

toScaler

  • toScaler(): number

  • Returns the only element.

    Returns number

    The only element

toString

  • toString(): string

  • Returns a string represented this tensor.

    Returns string

    String represented this tensor

transpose

  • transpose(...axises: (number | number[])[]): Tensor

  • Parameters

    • ...axises: (number | number[])[]

      Selected axises

    Returns Tensor

    Transposed tensor

StaticfromArray

  • fromArray(arr: number | Tensor | number[] | Matrix): Tensor

  • Returns a tensor from some value.

    Parameters

    Returns Tensor

    Tensor from some value

Staticones

  • ones(...size: number[]): Tensor

  • Returns a tensor filled with 1.

    Parameters

    • ...size: number[]

      Sizes for each dimension

    Returns Tensor

    Tensor filled with 1

  • ones(size: number[]): Tensor

  • Returns a tensor filled with 1.

    Parameters

    • size: number[]

      Sizes for each dimension

    Returns Tensor

    Tensor filled with 1

Staticrandn

  • randn(size: number[], myu?: number, sigma?: number): Tensor

  • Returns a tensor initialized normal random values.

    Parameters

    • size: number[]

      Sizes for each dimension

    • Optionalmyu: number

      Mean value of the Tensor

    • Optionalsigma: number

      Variance value of the Tensor

    Returns Tensor

    Tensor initialized normal random values

Staticrandom

  • random(size: number[], min?: number, max?: number): Tensor

  • Returns a tensor initialized uniform random values.

    Parameters

    • size: number[]

      Sizes for each dimension

    • Optionalmin: number

      Minimum value of the Tensor

    • Optionalmax: number

      Maximum value of the Tensor

    Returns Tensor

    Tensor initialized uniform random values

Staticzeros

  • zeros(...size: number[]): Tensor

  • Returns a tensor filled with 0.

    Parameters

    • ...size: number[]

      Sizes for each dimension

    Returns Tensor

    Tensor filled with 0

  • zeros(size: number[]): Tensor

  • Returns a tensor filled with 0.

    Parameters

    • size: number[]

      Sizes for each dimension

    Returns Tensor

    Tensor filled with 0

Settings

Member Visibility

On This Page

Constructors
Accessors
Methods