An Autoregressive Sequence Generator is a predictive model usually used for time-series based tasks in AI and ML. The operating principle of autoregressive models is predicated on using its own previously forecasted outputs as input data for following predictions. That is, the model predicts future sequence elements based on a linear function of previous elements. As such, these models are well-suited to tasks such as stock market predictions, energy consumption forecasting, weather forecasts, and any other applications involving sequential or temporal data.
Though autoregressive models initially found their applications in the field of economics and signal processing during the 1960s, it was not until the late 20th century, with increasing computational power and development in the field of AI and ML, that their use became more widespread.
The extension and implementation of autoregressive models in the AI and ML field are broad and many researchers have contributed significant work. However, some influential developments include the Autoregressive Integrated Moving Average (ARIMA) model introduced by George Box and Gwilym Jenkins in the 1970s, and more recently, attention to autoregressive models has been revived due to their integration within neural networks.