pypose.cumops

class pypose.cumops(input, dim, ops)

Returns the cumulative user-defined operation of LieTensor along a dimension.

\[y_i = x_1~\mathrm{\circ}~x_2 ~\mathrm{\circ}~ \cdots ~\mathrm{\circ}~ x_i, \]

where \(\mathrm{\circ}\) is the user-defined operation and \(x_i,~y_i\) are the \(i\)-th LieType item along the dim dimension of input and output, respectively.

Parameters:

• input (LieTensor) – the input LieTensor

• dim (int) – the dimension to do the operation over

• ops (func) – the user-defined operation or function

Returns:

LieTensor

Return type:

LieTensor

Note

• The users are supposed to provide meaningful operation.

• This function doesn’t check whether the results are valid for mathematical definition of LieTensor, e.g., quaternion.

• The time complexity of the function is \(\mathcal{O}(\log N)\), where \(N\) is the LieTensor size along the dim dimension.

Examples

>>> # The following operations are equivalent.
>>> input = pp.randn_SE3(2)
>>> input.cumprod(dim = 0, left=True)
SE3Type LieTensor:
tensor([[-0.6466,  0.2956,  2.4055, -0.4428,  0.1893,  0.3933,  0.7833],
        [ 1.2711,  1.2020,  0.0651, -0.0685,  0.6732,  0.7331, -0.0685]])
>>> pp.cumops(input, 0, lambda a, b : b @ a) # Left multiplication
SE3Type LieTensor:
tensor([[-0.6466,  0.2956,  2.4055, -0.4428,  0.1893,  0.3933,  0.7833],
        [ 1.2711,  1.2020,  0.0651, -0.0685,  0.6732,  0.7331, -0.0685]])