Cute 💙 Pets 😻🐶

Do you remember when you joined X? I do! #MyXAnniversary

Tao or Tau both work in a way

Master Mu? Master Tao?

If master foo is the character in Unix koan what is the name of the master in machine learning koans?

I spent all afternoon reading up on portable GUI frameworks that are lightweight, and have come to the conclusion we should just spin up a webserver...there is some wisdom in the #redbean project

I'm legit tired

There's a lot of GPT generated text going around, just wondering how good is it for being a tool for guided fiction writing? Any examples of people trying to use it in that way?

2/2 Contrary to image/text based DL models, tabular DL models need feature engineering to succeed (I know TabNet is a more recent thing...) Why isn't there more research on feature engineering for DL? Am I missing something?

1/2 After reading 2018 Airbnb's paper on Applying DL to Tabular data it boils down to: doing feature engineering to ensure features are between -1, 1 with median of 0 and "trying out" embeddings for categorical data.

4/4 Then we could created a federated "machine learning" framework that uses these (inefficient by Byzantine resilient) optimizers, covering linear models, tree models and more. Sure - it's not sexy, but it would hopefully be practical (I'm thinking like Vowpal Wabbit Library)

3/4 Stochastic Nelder-Mead can be used as a gradient free optimizer. It is also surprisingly Byzantine Resilient as it is insensitive to small inaccuracies, and leverages a centroid for its reflection (we can steal other variantions of the centroid to increase robustness)

1/4 - Byzantine Resilience/Optimization/Machine Learning Efficiency aside - why is Byzantine Resilience not discussed more broadly in the context of Optimization algorithms in Machine Learning? Is it because its not practical? No usecase? Difficult to productionize?

@k33g_org Thank you very much, I will take a look. Is there a curated list of "compatible packages" with TinyGo?

I'm surprised the CosmWasm/tinyjson repository has no stars or publicity even though it feels like the only project that properly enables TinyGo to have json support. Yes its "just a fork" of easyjson - I just thought there would be more effort from the community 🤷

@tier10k Looks action packed to me! Can't wait

2/n To get started, let's suppose you've built a model using your favorite system, and is supported by github.com/BayesWitnesses… - we can convert this to WebAssembly via Go with the power of TinyGo (of course you might just directly go to Typescript or Rust).

1/n #TinyML should be more of a thing! Let's talk about how to push models to WebAssembly! The future should be using interoperable formats that can be easily ported in a variety of programming languages and 100% portable. This is all made possible with the power of #wasmer

Is it even possible to learn development in the node/npm ecosystem with an old computer?

2/2 - but there is clearly a lack of impetus especially given Data Science/Machine Learning means a lack of resourcing besides companies that are truly looking to deploy on the edge. Hopefully WASM/WASI becomes the defacto standard to simplify all these pipelines