Hacker News Reader

GeoffNNon Timestral: Mistral release time-series foundation models

Not that I know of!

GeoffNNon Timestral: Mistral's time-series foundation models

omg - to keep the record straight, this person is not affiliated with The Forecasting Company

GeoffNNon Timestral: Mistral's time-series foundation models

Lots of friendly connections :) Both in Paris! Also same shape of model that we know well

GeoffNNon Timestral: Mistral release time-series foundation models

Not all of their trading is high freq

GeoffNNon Timestral: Mistral release time-series foundation models

Not dunking! Adding info/context for this field. We're very excited to see Mistral is building here too!

GeoffNNon Timestral: Mistral's time-series foundation models

Ah thanks! The Mistral model has a restrictive license, where you need a specific commercial agreement to use it if you make more than 20M$ revenue/month. Ours is Apache 2.

If you compare the 2, it's not a strict "timestral is better" story - the winrates are tight - your probably should try both on your use-case. We have a 42% winrate vs Timestral on fevbench for instance.

Ofc, our main value is not just these models, but also 1) extra capabilities (like our model that can natively account for stockouts/censored data, and others coming soon), 2) our systems and products around the model, and 3) we're all laser focused on time-series/forecasting

GeoffNNon Show HN: machine0 – Persistent NixOS VMs You Control from the CLI

Very cool! I'll definitely check this out for auto-research like experiments, esp. with GPUs

GeoffNNon A decoder-only foundation model for time-series forecasting

They say they can optionally use some time-dependent covariates (derived from the datetime) -- did anyone figure out if/how they are actually used in this model from the paper?