matrix v0.3.2 Matrix
Matrix is a linear algebra library for manipulating dense matrices. Its primary design goal is ease of use. It is desirable that the Matrix package interact with standard Elixir language constructs and other packages. The underlying storage mechanism is, therefore, Elixir lists.
A secondary design consideration is for the module to be reasonably efficient in terms of both memory usage and computations. Unfortunately there is a trade off between memory efficiency and computational efficiency. Where these requirements conflict Matrix will use the more computationally efficient algorithm.
Each matrix is represented as a “list of lists” whereby a 3x4 matrix is represented by a list of three items, with each item a list of 4 values. Constructors are provided for several common matrix configurations including zero filled, one filled, random filled the identity matrix, etc.
Examples
iex> Matrix.new(3, 4)
[[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]
iex> Matrix.ident(4)
[[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, 1]]
Summary
Functions
Returns a new matrix whose elements are the sum of the elements of the provided matrices. If the matrices are of differing sizes, the returned matrix will be the size and dimensions of the “overlap” between them. For instance, the sum of a 3x3 matrix with a 2x2 matrix will be 2x2. The sum of a 3x1 matrix with a 1x3 matrix will be 1x1
Compares two matrices as being (approximately) equal. Since floating point numbers have slightly different representations and accuracies on different architectures it is generally not a good idea to compare them directly. Rather numbers are considered equal if they are within an “epsilon” of each other. almost_equal compares all elements of two matrices, returning true if all elements are within the provided epsilon
Code borrowed from the ExMath library and duplicated here to reduce dependencies. ExMath is copyright © 2015 Ookami Kenrou ookamikenrou@gmail.com
Returns a new square “diagonal” matrix whose elements are zero except for the diagonal. The diagonal elements will be composed of the supplied list
Returns the value of the specified element (row and column) of the given
matrix (x). The row and column indices are zero-based. Returns default
if either row or col are out of bounds
Returns a new matrix whose elements are the element-by-element multiply of the elements of the provided matrices. Note that this is not the linear algebra matrix multiply. If the matrices are of differing sizes, the returned matrix will be the size and dimensions of the “overlap” between them. For instance, the element multiply of a 3x3 matrix with a 2x2 matrix will be 2x2, for a 3x1 matrix with a 1x3 matrix will be 1x1
Returns a new “identity” matrix of the specified size. The identity is defined as a square matrix with ones on the diagonal and zeros in all off-diagonal elements. Since the matrix is square only a single size parameter is required
Returns a new matrix which is the (linear algebra) inverse of the supplied matrix. If the supplied matrix is “x” then, by definition,
x * inv(x) = I
where I is the identity matrix. This function uses a brute force Gaussian elimination so it is not expected to be terribly fast
Returns the Kronecker tensor product of two matrices A and B. If A is an MxN and B is PxQ, then the returned matrix is an (MP)x(NQ) matrix formed by taking all possible products between the elements of A and the matrix B
Returns a new matrix which is the linear algebra matrix multiply
of the provided matrices. It is required that the number of columns
of the first matrix (x) be equal to the number of rows of the second
matrix (y). If x is an NxM and y is an MxP, the returned matrix product
xy is NxP. If the number of columns of x does not equal the number of
rows of y an ArgumentError
exception is thrown with the message “sizes
incompatible”
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with the supplied value “val” (default 0)
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with the ones
Returns a new matrix whose elements are identical to the supplied matrix x but with the supplied value appended to the end of each row
Returns a new matrix whose elements are identical to the supplied matrix x but with the supplied value appended to the beginning of each row
Returns a string which is a “pretty” representation of the supplied matrix
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with uniformly distributed random numbers between 0 and 1
Returns a new matrix whose elements are the elements of matrix x multiplied by the scale factor “s”
Returns a new matrix of the specified size (number of rows and columns) whose elements are sequential starting at 1 and increasing across the row
Returns a matrix that is a copy of the supplied matrix (x) with the specified element (row and column) set to the specified value (val). The row and column indices are zero-based. Negative indices indicate an offset from the end of the row or column. If an index is out of bounds, the original matrix is returned
Returns the size (dimensions) of the supplied matrix. The return value is a tuple of the dimensions of the matrix as {rows,cols}
Returns a new matrix whose elements are the difference (subtraction) of the elements of the provided matrices. If the matrices are of differing sizes, the returned matrix will be the size and dimensions of the “overlap” between them. For instance, the difference of a 3x3 matrix with a 2x2 matrix will be 2x2. The difference of a 3x1 matrix with a 1x3 matrix will be 1x1
Returns a new matrix whose elements are the transpose of the supplied matrix. The transpose essentially swaps rows for columns - that is, the first row becomes the first column, the second row becomes the second column, etc
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with the zeros
Types
Functions
Returns a new matrix whose elements are the sum of the elements of the provided matrices. If the matrices are of differing sizes, the returned matrix will be the size and dimensions of the “overlap” between them. For instance, the sum of a 3x3 matrix with a 2x2 matrix will be 2x2. The sum of a 3x1 matrix with a 1x3 matrix will be 1x1.
See also
Examples
iex> Matrix.add( Matrix.ident(3), Matrix.ident(3) )
[[2, 0, 0], [0, 2, 0], [0, 0, 2]]
iex> Matrix.add( Matrix.ones(3,3), Matrix.ones(2,2) )
[[2, 2], [2, 2]]
iex> Matrix.add( Matrix.ones(3,1), Matrix.ones(1,3) )
[[2]]
Compares two matrices as being (approximately) equal. Since floating point numbers have slightly different representations and accuracies on different architectures it is generally not a good idea to compare them directly. Rather numbers are considered equal if they are within an “epsilon” of each other. almost_equal compares all elements of two matrices, returning true if all elements are within the provided epsilon.
Examples
iex> Matrix.almost_equal( [[1, 0], [0, 1]], [[1,0], [0,1+1.0e-12]] )
false
iex> Matrix.almost_equal( [[1, 0], [0, 1]], [[1,0], [0,1+0.5e-12]] )
true
close_enough?(number, number, number, non_neg_integer) :: boolean
Code borrowed from the ExMath library and duplicated here to reduce dependencies. ExMath is copyright © 2015 Ookami Kenrou ookamikenrou@gmail.com
Equality comparison for floating point numbers, based on this blog post by Bruce Dawson.
Returns a new square “diagonal” matrix whose elements are zero except for the diagonal. The diagonal elements will be composed of the supplied list
See also
Examples
iex> Matrix.diag([1,2,3])
[[1, 0, 0], [0, 2, 0], [0, 0, 3]]
Returns the value of the specified element (row and column) of the given
matrix (x). The row and column indices are zero-based. Returns default
if either row or col are out of bounds.
See also
Examples
iex> Matrix.elem( Matrix.ident(3), 0,0 )
1
Returns a new matrix whose elements are the element-by-element multiply of the elements of the provided matrices. Note that this is not the linear algebra matrix multiply. If the matrices are of differing sizes, the returned matrix will be the size and dimensions of the “overlap” between them. For instance, the element multiply of a 3x3 matrix with a 2x2 matrix will be 2x2, for a 3x1 matrix with a 1x3 matrix will be 1x1.
See also
Examples
iex> Matrix.emult( Matrix.new(3,3,2), Matrix.new(3,3,-2) )
[[-4, -4, -4], [-4, -4, -4], [-4, -4, -4]]
iex> Matrix.emult( Matrix.ones(3,3), Matrix.ones(2,2) )
[[1, 1], [1, 1]]
iex> Matrix.emult( Matrix.ones(3,1), Matrix.ones(1,3) )
[[1]]
Returns a new “identity” matrix of the specified size. The identity is defined as a square matrix with ones on the diagonal and zeros in all off-diagonal elements. Since the matrix is square only a single size parameter is required.
See also
Examples
iex> Matrix.ident(3)
[[1, 0, 0], [0, 1, 0], [0, 0, 1]]
Returns a new matrix which is the (linear algebra) inverse of the supplied matrix. If the supplied matrix is “x” then, by definition,
x * inv(x) = I
where I is the identity matrix. This function uses a brute force Gaussian elimination so it is not expected to be terribly fast.
Examples
iex> x = Matrix.rand(5,5) iex> res = Matrix.mult( x, Matrix.inv(x) ) iex> Matrix.almost_equal(res,[[1,0,0],[0,1,0],[0,0,1]]) true
Returns the Kronecker tensor product of two matrices A and B. If A is an MxN and B is PxQ, then the returned matrix is an (MP)x(NQ) matrix formed by taking all possible products between the elements of A and the matrix B.
A = |1000| B = | 1 -1| |0100| | -1 1| |0010| |0001|then
kron(A,B) = | 1 -1 0 0 0 0 0 0| | -1 1 0 0 0 0 0 0| | 0 0 1 -1 0 0 0 0| | 0 0 -1 1 0 0 0 0| | 0 0 0 0 1 -1 0 0| | 0 0 0 0 -1 1 0 0| | 0 0 0 0 0 0 1 -1| | 0 0 0 0 0 0 -1 1|
Returns a new matrix which is the linear algebra matrix multiply
of the provided matrices. It is required that the number of columns
of the first matrix (x) be equal to the number of rows of the second
matrix (y). If x is an NxM and y is an MxP, the returned matrix product
xy is NxP. If the number of columns of x does not equal the number of
rows of y an ArgumentError
exception is thrown with the message “sizes
incompatible”
See also
Examples
iex> Matrix.mult( Matrix.seq(2,2), Matrix.seq(2,2) )
[[7, 10], [15, 22]]
iex> Matrix.mult( Matrix.ones(3,2), Matrix.ones(2,3) )
[[2, 2, 2], [2, 2, 2], [2, 2, 2]]
iex> Matrix.mult( Matrix.ones(3,2), Matrix.ones(3,2) )
** (ArgumentError) sizes incompatible
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with the supplied value “val” (default 0).
See also
Examples
iex> Matrix.new(3, 4)
[[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]
iex> Matrix.new(2, 3, -10)
[[-10, -10, -10], [-10, -10, -10]]
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with the ones.
See also
Examples
iex> Matrix.ones(3, 4)
[[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]]
Returns a new matrix whose elements are identical to the supplied matrix x but with the supplied value appended to the end of each row.
See also
Examples
iex> Matrix.postfix_rows( Matrix.seq(2,2), 10 )
[[1, 2, 10], [3, 4, 10]]
Returns a new matrix whose elements are identical to the supplied matrix x but with the supplied value appended to the beginning of each row.
See also
Examples
iex> Matrix.prefix_rows( Matrix.seq(2,2), 10 )
[[10, 1, 2], [10, 3, 4]]
Returns a string which is a “pretty” representation of the supplied matrix.
Returns a new matrix of the specified size (number of rows and columns). All elements of the matrix are filled with uniformly distributed random numbers between 0 and 1.
Examples
iex> _ = :rand.seed(:exs1024, {123, 123534, 345345})
iex> Matrix.rand(3,3)
[[0.5820506340260994, 0.6739535732076178, 0.9178030245386003],
[0.7402049520743949, 0.5589108995145826, 0.8687305849540213],
[0.8851580858928109, 0.988438251464987, 0.18105169154176423]]
See also
Returns a new matrix whose elements are the elements of matrix x multiplied by the scale factor “s”.
Examples
iex> Matrix.scale( Matrix.ident(3), 2 )
[[2,0,0], [0,2,0], [0,0,2]]
iex> Matrix.scale( Matrix.ones(3,4), -2 )
[[-2, -2, -2, -2], [-2, -2, -2, -2], [-2, -2, -2, -2]]
Returns a new matrix of the specified size (number of rows and columns) whose elements are sequential starting at 1 and increasing across the row.
See also
new/3, ones/2, rand/2, zeros/2
Examples
iex> Matrix.seq(3,2)
[[1, 2], [3, 4], [5, 6]]
Returns a matrix that is a copy of the supplied matrix (x) with the specified element (row and column) set to the specified value (val). The row and column indices are zero-based. Negative indices indicate an offset from the end of the row or column. If an index is out of bounds, the original matrix is returned.
See also
Examples
iex> Matrix.set( Matrix.ident(3), 0,0, -1)
[[-1, 0, 0], [0, 1, 0], [0, 0, 1]]
Returns the size (dimensions) of the supplied matrix. The return value is a tuple of the dimensions of the matrix as {rows,cols}.
See also
Examples
iex> Matrix.size( Matrix.new(3,4) )
{3, 4}
Returns a new matrix whose elements are the difference (subtraction) of the elements of the provided matrices. If the matrices are of differing sizes, the returned matrix will be the size and dimensions of the “overlap” between them. For instance, the difference of a 3x3 matrix with a 2x2 matrix will be 2x2. The difference of a 3x1 matrix with a 1x3 matrix will be 1x1.
See also
Examples
iex> Matrix.sub( Matrix.ident(3), Matrix.ones(3,3) )
[[0, -1, -1], [-1, 0, -1], [-1, -1, 0]]
iex> Matrix.sub( Matrix.ones(3,3), Matrix.ones(2,2) )
[[0, 0], [0, 0]]
iex> Matrix.sub( Matrix.ones(3,1), Matrix.ones(1,3) )
[[0]]
Returns a new matrix whose elements are the transpose of the supplied matrix. The transpose essentially swaps rows for columns - that is, the first row becomes the first column, the second row becomes the second column, etc.
Examples
iex> Matrix.transpose( Matrix.seq(3,2) ) [[1, 3, 5], [2, 4, 6]]