FLAML VS stan

I really like the simplicity of this framework, and they hit on a lot of common problems found in other agent-based frameworks. Most intrigued by the RAG improvements.

Seems like Microsoft was frustrated with the pace of movement in this space and the shitty results of agents (which admittedly kept my interest turned away from agents for the last few months). I'm interested again because it makes practical sense, and from looking at the example notebooks, seems fairly easy to integrate into existing applications.

Maybe this is the 'low code' approach that might actually work, and bridge together engineering and non-engineering resources.

This example was what caught my eye: https://github.com/microsoft/FLAML/blob/main/notebook/autoge...

4. FLAML

Improved AutoML - Again, with collaboration from Microsoft Research, we used FLAML to update our existing AutoML solutions. What does this mean for you? You're using the latest techniques but all you need is a problem to solve and some data to get started.

For a complete tutorial, navigate to this Jupyter Notebook: https://github.com/microsoft/FLAML/blob/main/notebook/flaml_automl.ipynb

Microsoft Researchers have developed FLAML (Fast Lightweight AutoML) which can now utilize Ray Tune for distributed hyperparameter tuning to scale up FLAML’s resource-efficient & seamlessly parallelizable algorithms across a cluster.

Looks nice but I wonder if this is practical for non-tiny problems. The papers are a bit hard to follow but it looks like training is restarted with every new architecture choice. As for the library itself, the only large neural net example is a finetune of an NLP model that only searches over ADAM's optimizer params - which could be useful but it's a stretch to call that AutoML.

Alternatives: stan and edward

remove.packages(c("StanHeaders", "rstan")) install.packages("rstan", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))

My approach to problems like this is to write down the proposed model mathematically first, in extreme detail. I find hierarchical form to be the easiest way to break it down piece by piece. Once I have the maths then I turn it into a Stan model. Last step is to use the Stan output to answer the research questions.

Here is a good breakdown of recommendations from Andrew Gelman.

For instance my first choice in these cases is always a Bayesian inference tool like Stan. In my experience as someone who’s more of a programmer than mathematician/statistician, Bayesian tools like this make it much easier to not accidentally fool yourself with assumptions, and they can be pretty good at catching statistical mistakes.

I tend to be most impressed by tools and libraries. The stuff that has most impressed me in my time in ML is stuff like pytorch and Stan, tools that allow expression of a wide variety of statistical (and ML, DL models, if you believe there's a distinction) models and inference from those models. These are the things that have had the largest effect in my own work, not in the sense of just using these tools, but learning from their design and emulating what makes them successful.

oh its certainly used in practice. you should look into frameworks like Stan[1] and pyro[2]. i think bayesian models are seen as more explainable so they will be used in industries that value that sort of thing

[1] https://mc-stan.org/