阅读笔记DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks
简介DeepAR是一个概率预测模型，给历史数据，预测的是未来的分布。方法zi,t∈Rz_{i,t}\in\mathbb{R}zi,t∈R表示时间序列iii在ttt时刻的值。给一个连续时间段t∈[1,T]t\in[1,T]t∈[1,T]，将其划分为contextwindow[1,t0)[1,t_0)[1,t0)和predictionwindow[t0,T][t_0,T][t0,T]。用context