Join Camilo Santo, Machine Learning Engineer at Bearing.AI, as he discusses new release TimeGPT with Azul Garza Ramírez, CTO and Co-founder of Nixtla, creators of TimeGPT. They will evaluates how it stacks up against other time series models.
Here are the Notebook ( colab.research.google.com/dri...) and Data (drive.google.com/file/d/1A5NO....
Here are some answers to questions we didn't get to during the event:
Q: Can TimeGPT be extended for time series classification?
A: Currently we only support forecasting and anomaly detection. we are improving TimeGPT to have classification capabilities in the next releases.
Q: TimeGPT is an API only now - are there any plans to make that open source?
A: Yes. we have plans to make some versions open source. You can follow our open-source organization (github.com/Nixtla) or social networks ( / nixtlainc ) to stay tuned.
Q: What kind of data was TimeGPT trained on?
A: TimeGPT was trained on more than 100B data points including finance (some stock data), retail, electricity, among others.
Q: Apart from anomalies and historical forecasting, what other use cases does TImeGPT have API calls for?
A: Currently, we support anomaly detection and forecasting, including multivariate forecasting (external regressors), uncertainty quantification, fine-tuning, forecasting multiple series at once, and the addition of holidays and special dates, among other exciting features that you can find in our official documentation (docs.nixtla.io/docs/timegpt_q....
Негізгі бет Ғылым және технология How does TimeGPT stack up?
Пікірлер: 6