Nico Szerman

Msgpack looks amazing! I think that's the way to go. In our case, I think Msgpack is better than gzipping because of streaming unpacking. When dealing with massive datasets we...

Probably the best way to approach this problem is by treating it as an exploration game in which the NN moves through the graph, time and wrong guesses are penalized,...

With a variation of [Graph Convolutional Networks](https://towardsdatascience.com/how-to-do-deep-learning-on-graphs-with-graph-convolutional-networks-7d2250723780), we could come up with a cool model that works for graphs of any size. I'm also interested in whether the simple model...

We're planning to start supporting I/O of any format after an upcoming refactor (eta probably like two weeks or less). Until then, probably the best is to just destructure the...

I'm trying to understand your idea. How would the neural network process the input obiect?

Makes sense! I like your idea. We need to think if it's best to add this functionality under .train() or if we should add it under another layer of abstraction....

By checking that it's a function, we know that when running eval on cost.toString() only a function will be returned. This is still dangerous though. Maybe the way forward is...

There's only one eval() call in the file On Wed, Jun 26, 2019, 2:00 PM Christian Echevarria wrote: > @nicszerman can you post some code > snippets so we can...

Thank you! Those are the requirements I was looking for. So with minimal users any modern laptop should be fine? Are 512 MB ram enough? If you do list the...

Thank you. Was wondering about the 512 MB because that’s the typical “free tier” server hardware on hosting providers. I’m considering running Estuary in the same server as the main...