textremo-toolbox 1.0.16

No project description provided

Toolbox

This repositories offers a toolbox across platforms.

How to use

• MatlabFuncs: this class simulate all Matlab functions for numpy, pytorch, tensorflow. Please note that data must be batched (at least batch_size=1).
• MatlabFuncHelper: this class simulate all Matlab functions in python (batch is supported).
  • batch
    • setBatchSize(batch_size) or setBS(batch_size)
    • getBatchSize() or getBS()
  • data type selectors
    • toNP(): to numpy
    • toPT(): to pytorch
    • toTF(): to tensorflow
  • checkers
    • isvector(mat)
    • ismatrix(mat)
    • isnan(mat)
  • generators
    • zeros(d1, d2, ...)
    • ones(d1, d2, ...)
    • eye(d1, ...): eye(d1) or eye(d1, d2)
    • seq(): generate a sequence, seq(end), seq(beg, end), seq(beg, step, end)
    • rand(d1, d2, ...):
    • randn(d1, d2, ...):
    • dftmtx(d1): generate the discrete transform matrix
  • transformers
    • squeeze(mat): remove redundant dimension except for the batch_size
    • reshape(mat, d1, d2, ...):
    • repmat1(mat, d1, d2, ...): repeat the matrix in the given dimension (the batch dim is repeated as 1)
    • repmatN(mat, d1, d2, ...): repeat the matrix in the given dimension (the batch dim is repeated as batch_size)
    • diag(mat): generate a matrix based on its diag or get a diagonal matrix from its vector
      @mat: a vector as [(batch_size), n, 1] or [(batch_size), 1, n]; if n == 1, it will be taken as [(batch_size), n, 1]. Or a square matrix [(batch_size), n, n]
    • circshift(mat, step): circular shift (1st index except for the batch size)
  • Maths
    • sum(): sum along the last axis
    • max(): return the maximum of a matrix or the maximum of two matrices (for complex value, we compare the magnitude)

      self.max(mat1, axis=-1);    // return the maximum of a matrix, axis tells which axis to look at
      self.max(mat1, 4);          // return the maximum of a matrix and the given value
      self.max(mat1, mat2);       // return the maximum of two matrices
      

    • kron(a, b): Kronecker produc