Transfer learning is proven to be an effective way to incrementally learn with limited data, mostly in image applications. However, there are cases where only time series data are available.
The purpose of this project is to investigate transfer learning for crop time series data. As a first step, the student will experiment with different topologies of neural networks to predict crop yield from time series data. A topology providing sufficiently good results, based on predefined metrics, will be selected. Then, a smaller dataset with another aspect affecting crop yield will be provided. The student will train the selected neural network with the new dataset and report the results.
- Prepare/process timeseries data for machine learning
- Develop a deep learning pipeline
- Evaluate the deep learning approach in a case study
- Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. The MIT Press.
- Pan, S. J., & Yang, Q. (2009). A survey on transfer learning. IEEE Transactions on Knowledge and Data Engineering, 22(10), 1345–1359.
- Sun, Q., Liu, Y., Chua, T.-S., & Schiele, B. (2019). Meta-transfer learning for few-shot learning. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 403–412.
- FTE-35306 Machine Learning
- GRS-34806 Deep Learning
Theme(s): Modelling & visualisation