Microsoft Excel becomes a deep learning sandbox with The Architect
Turn Excel into an interactive deep learning lab with live matrix math and backpropagation.
Researchers Mohammad Imrul Jubair and Tom Yeh present The Architect, a system that turns Microsoft Excel into an interactive view of deep learning mathematics. A user describes a neural network in a compact table, and the system generates a workbook showing the full forward pass—and, when requested, the backward pass and parameter updates—as live spreadsheet formulas. Inputs, weights, labels, and hyperparameters stay editable, with Excel reactively updating dependent computations. It also produces aligned PyTorch snippets to help connect formulas to implementation.
- The Architect turns Excel into an interactive deep learning lab, exposing forward/backward passes, gradients, and updates as live formulas
- Users define models in a compact table; system generates Excel workbooks with reactive computations and aligned PyTorch code
- Supports debugging dying ReLU, vanishing gradients, and learning rate exploration with inspectable matrix math
Why It Matters
Finally, a way to see the 'black box' of deep learning—with real numbers you can tweak in a familiar tool.