Applied timeseries Transfer learning

Nikolay Laptev, Jiafan Yu, Ram Rajagopal

Abstract:

Reliable and accurate time-series modeling is critical in many fields including energy, finance, and manufacturing.
Many time-series tasks, however, suffer from a limited amount of clean training data resulting in poor forecasting, classification or clustering performance. Recently, convolutional neural networks (CNNs) have shown outstanding image classification performance even on tasks with small-scale training sets. The performance can be attributed to transfer learning through ability of CNNs to learn rich mid-level image representations. For time-series, however, no prior work exists on general transfer learning. In this short paper, motivated by recent success of transfer learning in image-related tasks, we are the first to show that using an LSTM auto-encoder with attention trained on a large-scale timeseries dataset with pre-processing we can effectively transfer time-series features across diverse domains.

TL;DR: transfer learning in time-series

Keywords: time-series, transfer-learning