This is the recommended learning path for VTL. It starts with tensors, then builds toward autograd, neural networks, optimizers, and GPU-backed examples.
- First Steps — creation, indexing, shapes.
- Slicing — views and sub-tensors.
- Broadcasting — shape-compatible operations.
- Map and Reduce — element-wise transforms and reductions.
- Reductions — argmax, argmin, cumulative operations.
- Matrix and Vector operations — VSL-backed LA.
- Advanced Linear Algebra — QR, LU, Cholesky, pinv.
- Automatic Differentiation —
Variable, gates, backprop. - Neural Networks — layers, losses,
Sequential. - Optimizers — SGD, Adam, AdamW, schedulers.