MANATEE(lm): Market Analysis based on language model architectures | Hacker News Reader

MANATEE(lm): Market Analysis based on language model architectures

colab.research.google.com

31 pointsbrulenaudet9 comments

I have opened-sourced a side-project called MANATEE(lm): Market Analysis based on language model architectures, as a test case for the amazon/chronos-t5-large model from Amazon Web Services (AWS), UC San Diego and Albert-Ludwigs-Universität Freiburg, hosted on Hugging Face, in a Google Colab notebook.

Using Polars for parallelized operations, enabling concurrent execution of data transformations and computations across multiple CPU cores, and Plotly for data plotting and visualization, this notebook may be an entry point for time series analysis on various domains, including retail, energy, finance, healthcare, climate science.

(edit) needs clarification that this is a personal example, not a product from Google Research

I wouldn't assume that's intentional on the poster's part. HN should probably just add a rule for Colab notebooks as another top-level comment to this thread pointed out.

This isn’t my field, but possibly relevant:

Why Doesn’t My Model Work? https://thegradient.pub/why-doesnt-my-model-work/

It describes a lot of pitfalls, some specific to time-series forecasting, and it links to this paper:

Forecast evaluation for data scientists: common pitfalls and best practices https://link.springer.com/article/10.1007/s10618-022-00894-5

I'm not seeing anything resembling an evaluation of model performance on a quick scroll.

The HN-parsed URL makes it looks like it is from Google (research.google.com), but this is a Colab notebook written by someone who doesn't work at Google. I think HN needs to include the colab.research.google.com on all Colab notebooks to better inform HN readers.

The only reason I clicked on a post about market analysis with LLMs is because I thought it was a project from Google. This is a notoriously difficult area with a lot of BS, and LLMs only raise my BS meter. So I thought to myself that if Google Research feels like they did something they want to share in this space, I might have to reconsider and check it out.

I was so disappointed. The notebook felt unfinished.

Agreed, I have no idea what the plot at the end means and most outputs are just tensors

Less of a HN issue and more google allowing UGC on their research subdomain

Reply on news.ycombinator.com