Demo: DeepONet: Einstein Summation
Interactive DeepONet Einsum Demo
Explore how torch.einsum('bp,bnp->bn', branch_out, trunk_out) combines branch and trunk network outputs step by step.
DeepONet Einsum Operation
DeepONet Forward Pass:
result = torch.einsum('bp,bnp->bn', branch_out, trunk_out)
This computes dot products between branch features and trunk features for each query point
Mathematical Operation:
result[b,n] = Σ(p=0 to P-1) branch[b,p] × trunk[b,n,p]
For each batch b and query point n, sum over all feature dimensions p
In DeepONet: Branch learns function representations, Trunk learns spatial/temporal coordinates.
Einsum combines them to predict function values at any query location.
DeepONet Architecture with Batching
Tensor Flow Visualization
Step-by-Step Computation
Branch Output Vectors [batch, p]
Trunk Feature Matrix [batch, n, p]
Final Result [batch, n]
Step-by-Step Computation Details
©
|
Cornell University
|
Center for Advanced Computing
|
Copyright Statement
|
Access Statement
CVW material development is supported by NSF OAC awards 1854828, 2321040, 2323116 (UT Austin) and 2005506 (Indiana University)
CVW material development is supported by NSF OAC awards 1854828, 2321040, 2323116 (UT Austin) and 2005506 (Indiana University)