ELIGIBILITY: This project is for a student taking the MSc in Statistics and Sustainability. Interested students should contact Gabriel Palma at GPALMA@tcd.ie
Deep Learning (DL)-based clustering methods, such as Variational Deep Embeddings (VaDE) [1], combine model‑based clustering (Gaussian Mixture Models, GMMs) with variational autoencoders and have shown promising results on clustering benchmarks.
Building on recent advances in DL clustering and the crucial role of clustering in classifying insect outbreaks with the Pattern‑Based Prediction (PBP) algorithm [2], this project will explore novel DL‑based clustering techniques to improve PBP for predicting insect‑pest outbreaks in real‑world datasets. The student will work directly with the PyPBP package and evaluate how different DL architectures affect the clustering stage of the model.
A familiarity with Python programming language, model based clustering and concepts of deep learning are expected from the student.
References
[1] Zhuxi Jiang, Yin Zheng, Huachun Tan, Bangsheng Tang, Hanning Zhou, Variational deep embedding: An unsupervised and generative approach to clustering, Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17), 2017, pp. 1965–1972, https://doi.org/10.24963/ijcai.2017/273.
[2] Gabriel R. Palma, Wesley A.C. Godoy, Eduardo Engel, Douglas Lau, Edgar Galvan, Oliver Mason, Charles Markham, Rafael A. Moral, Pattern-based prediction of population outbreaks, Ecological Informatics, Volume 77, 2023, 102220, ISSN 1574-9541, https://doi.org/10.1016/j.ecoinf.2023.102220.
[1] Zhuxi Jiang, Yin Zheng, Huachun Tan, Bangsheng Tang, Hanning Zhou, Variational deep embedding: An unsupervised and generative approach to clustering, Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17), 2017, pp. 1965–1972, https://doi.org/10.24963/ijcai.2017/273.
[2] Gabriel R. Palma, Wesley A.C. Godoy, Eduardo Engel, Douglas Lau, Edgar Galvan, Oliver Mason, Charles Markham, Rafael A. Moral, Pattern-based prediction of population outbreaks, Ecological Informatics, Volume 77, 2023, 102220, ISSN 1574-9541, https://doi.org/10.1016/j.ecoinf.2023.102220.