"Just posted “Ludwig: a type-based declarative deep learning toolbox”, a paper describing the design of #Ludwig and the motivations behind its architecture: <a href="https://arxiv.org/abs/1909.07930"" rel="nofollow">https://arxiv.org/abs/1909.07930"</a><p><a href="https://twitter.com/w4nderlus7/status/1174244681003208704" rel="nofollow">https://twitter.com/w4nderlus7/status/1174244681003208704</a>
Related discussion from the original release of Ludwig: <a href="https://news.ycombinator.com/item?id=19141001" rel="nofollow">https://news.ycombinator.com/item?id=19141001</a>