3 ms·

Chronos: Pretrained (Language) Models for Probabilistic Time Series Forecasting

• cs702 3y ago
  TL;DR: * Scale the time series data and quantize the floating point values into B bins. * Each bin becomes a corresponding token id in a vocabulary of B embeddings. * Train a small LLM to predict the next token id given a sequence of token ids. * At each time step, the LLM gives you a probability distribution over B bins.
• huibin_shen 3y ago
  On a large scale of 42 time series datasets, Chronos demonstrates impressive empirical performance. In the zero-shot setting, it matches or even outperforms many baselines which are trained on the dataset.
• sebastianpineda 3y ago
  this is an amazing contribution for the community!