Intra-domain and cross-domain transfer learning for time series data -- How transferable are the features?

TL;DR

This paper investigates the effectiveness of transfer learning across different time series domains, assessing its impact on model performance and convergence with small datasets, through extensive experiments and statistical validation.

Contribution

It provides a comprehensive analysis of intra- and cross-domain transfer learning for time series data, identifying conditions that influence transfer effectiveness and offering insights into domain compatibility.

Findings

Transfer learning generally improves or maintains model performance.

Cross-domain transfer can be effective depending on domain similarity.

Target dataset size and model hyperparameters influence transfer success.

Abstract

In practice, it is very demanding and sometimes impossible to collect datasets of tagged data large enough to successfully train a machine learning model, and one possible solution to this problem is transfer learning. This study aims to assess how transferable are the features between different domains of time series data and under which conditions. The effects of transfer learning are observed in terms of predictive performance of the models and their convergence rate during training. In our experiment, we use reduced data sets of 1,500 and 9,000 data instances to mimic real world conditions. Using the same scaled-down datasets, we trained two sets of machine learning models: those that were trained with transfer learning and those that were trained from scratch. Four machine learning models were used for the experiment. Transfer of knowledge was performed within the same domain of application (seismology), as well as between mutually different domains of application (seismology, speech, medicine, finance). We observe the predictive performance of the models and the convergence rate during the training. In order to confirm the validity of the obtained results, we repeated the experiments seven times and applied statistical tests to confirm the significance of the results. The general conclusion of our study is that transfer learning is very likely to either increase or not negatively affect the predictive performance of the model or its convergence rate. The collected data is analysed in more details to determine which source and target domains are compatible for transfer of knowledge. We also analyse the effect of target dataset size and the selection of model and its hyperparameters on the effects of transfer learning.