Advanced Examples

[srush]

Some more classic ML examples that might be interesting:

• [ ] Graph Neural Networks (message passing) with the sparse indices
• [ ] Optimizers (ADAM / AdaGrad, not sure if they are interesting to do)
• [ ] Topic modeling (LDA complex indexed updates)
• [ ] Deep Boltzman Machine (MCMC over many variables)
• [ ] Kernels machines (Some of these get a bit dimensionally complex)

[davidweichiang]

• [ ] Belief propagation (@ctongfei has written text that has not been incorporated yet) or some other extended example illustrating how names = random variables.
• [ ] More about Gaussians (@srush)

[davidweichiang]

Optimizers would become interesting IMO only if we had an extension to sets of tensors, like

\theta = {W, b} \theta \leftarrow \theta + \eta \partial f / \partial\theta