Machine learning and other gibberish
See also: https://sharing.leima.is
Archives: https://datumorphism.leima.is/amneumarkt/
#dl

Google & USC benchmarked a prompt based forecasting method, and the results are amazing.

Cao D, Jia F, Arik SO, Pfister T, Zheng Y, Ye W, et al. TEMPO: Prompt-based Generative Pre-trained Transformer for time series forecasting. arXiv [cs.LG]. 2023. Available: http://arxiv.org/abs/2310.04948
#misc

Thinking about the past few days, I realized I never used stackoverflow during work. Our company pays for our github copilot services so I completely switched to the copilot chats.
So sad but I think I can totally live in a world without stackoverflow nowadays.

https://www.theverge.com/2023/10/16/23919004/stack-overflow-layoff-ai-profitability Stack Overflow lays off over 100 people as the AI coding boom continues
#llm

You probably heard of the famous ChatDev (https://github.com/OpenBMB/ChatDev ) that works like a whole company with all sorts of employees solving different parts of a complicated problem. If not, checkout the video demo in the GitHub repo.

Now, Microsoft joined the game. They built a framework for such models. The screenshot attached is a demo from the documentation. I tried it, and it is super easy to get started.

https://github.com/microsoft/autogen
#misc

https://avi-loeb.medium.com/we-have-discovered-spherules-from-the-path-of-the-first-recognized-interstellar-meteor-im1-d6cd94946b53

When I was doing my PhD, I spent a lot of hours on Loeb's papers. I was always hoping to see some evidence that there is already a party going on in the universe but we, as human beings, haven't found the way to chime in. I even wrote some stories on this idea, e.g.,
https://blog.leima.is/stories/dream-a-new-world/

I don't know. It is quite shameful to believe in something like this as an adult. But then again, who knows. At least I can have some fun. We Have Discovered Spherules from the Path of the First Recognized Interstellar Meteor, IM1
#ml

A family tree shows how transformers are evolving.

(HTML is probably the worst name for a model.)

https://arxiv.org/abs/2302.07730
#forecasting

Created some sections on forecasting with trees. This first draft provides some first steps to applying trees to forecasting problems as well as some useful theories about tree-based models.

https://dl.leima.is/trees/tree.basics/

https://dl.leima.is/notebooks/tree_darts_random_forest/
Back to Top