PyLops-GPU API¶
Linear operators¶
Templates¶
LinearOperator(shape, dtype[, Op, explicit, …]) |
Common interface for performing matrix-vector products. |
TorchOperator(Op[, batch, pylops, device]) |
Wrap a PyLops operator into a Torch function. |
Basic operators¶
MatrixMult(A[, dims, device, togpu, tocpu, …]) |
Matrix multiplication. |
Identity(N[, M, inplace, complex, device, …]) |
Identity operator. |
Diagonal(diag[, dims, dir, device, togpu, …]) |
Diagonal operator. |
VStack(ops[, device, togpu, tocpu, dtype]) |
Vertical stacking. |
Smoothing and derivatives¶
FirstDerivative(N[, dims, dir, sampling, …]) |
First derivative. |
SecondDerivative(N[, dims, dir, sampling, …]) |
Second derivative. |
Laplacian(dims[, dirs, weights, sampling, …]) |
Laplacian. |
Signal processing¶
Convolve1D(N, h[, offset, dims, dir, …]) |
1D convolution operator. |
Solvers¶
Low-level solvers¶
cg(A, y[, x, niter, tol]) |
Conjugate gradient |
cgls(A, y[, x, niter, damp, tol]) |
Conjugate gradient least squares |
Least-squares¶
leastsquares.NormalEquationsInversion(Op, …) |
Inversion of normal equations. |
Sparsity¶
sparsity.FISTA(Op, data, niter[, eps, …]) |
Fast Iterative Soft Thresholding Algorithm (FISTA). |
sparsity.SplitBregman(Op, RegsL1, data[, …]) |
Split Bregman for mixed L2-L1 norms. |
Applications¶
Geophysical subsurface characterization¶
poststack.PoststackInversion(data, wav[, …]) |
Post-stack linearized seismic inversion. |