3 ms·
Show HN: Spektral – Deep learning on graphs with Keras
- mfazalul 8y agoAs soon as I saw the title, my reaction was "YESSSS!"... A while ago, I was looking for something like this, but I soon realized that deep learning on graphs is still comparatively new, and not many python modules exist for it.
- Phait 8y agoGlad you like it. Now go and build crazy stuff!
- helloiloveyou 8y agoYEEES I will finally be able to complete a college assignment that I postponed
- syntaxing 8y agoHow is the data fed into the model? Are the graphs represented by a dictionary or an adjacency matrix (or neither)? Does anyone have more links I can read more about graph based deep learning?
- mfazalul 8y agoThe official docs have some of the details: https://danielegrattarola.github.io/spektral/data/ https://danielegrattarola.github.io/spektral/data/ More about Graph NNs: https://arxiv.org/pdf/1812.08434.pdf https://arxiv.org/pdf/1812.08434.pdf
- syntaxing 8y agoThank you!!
- joebarbere 8y agoMachine Learning in Complex Networks https://link.springer.com/book/10.1007/978-3-319-17290-3 https://link.springer.com/book/10.1007/978-3-319-17290-3 Front Matter - https://link.springer.com/content/pdf/bfm%3A978-3-319-17290-3%2F1.pdf https://link.springer.com/content/pdf/bfm%3A978-3-319-17290-...
- sjg007 8y agoVectorize the adjacency matrix.
- JoeDaDude 8y agoI'd love to tinker with this. What is a classic problem or demo for graph neural networks? Is there an equivalent to the MNIST problem for GNNs that would make a good learning exercise?