Iterative self-transfer learning: A general methodology for response time-history prediction based on small dataset
This addresses a bottleneck in surrogate modeling for engineering applications where data scarcity is common, though it appears incremental as it builds on transfer learning and pseudo-label strategies.
The paper tackles the problem of response time-history prediction with limited data by proposing an iterative self-transfer learning method, which improves model performance by nearly an order of magnitude on small datasets without requiring external labeled samples or pre-trained models.
There are numerous advantages of deep neural network surrogate modeling for response time-history prediction. However, due to the high cost of refined numerical simulations and actual experiments, the lack of data has become an unavoidable bottleneck in practical applications. An iterative self-transfer learningmethod for training neural networks based on small datasets is proposed in this study. A new mapping-based transfer learning network, named as deep adaptation network with three branches for regression (DAN-TR), is proposed. A general iterative network training strategy is developed by coupling DAN-TR and the pseudo-label strategy, and the establishment of corresponding datasets is also discussed. Finally, a complex component is selected as a case study. The results show that the proposed method can improve the model performance by near an order of magnitude on small datasets without the need of external labeled samples,well behaved pre-trainedmodels, additional artificial labeling, and complex physical/mathematical analysis.