ipab workshop-09-11-2023

Speaker: Alexandros Keros

Title: Topological methods for learning and guiding self-assembly of nanoparticles

Abstract: Nanoparticles have the phenomenal ability to self-assemble into patterned surfaces and specialised nanostructures. As their form dictates function, the opportunity arises for top-down synthesis of functionalized nanomaterials, with widespread impact in biomedical, electronics, and energy conversion applications. While pioneering experimental work showcases the potentialities of guiding self-assembly, fabricating such materials is time consuming, and requires a fair amount of expert intuition. Similarly, particle simulations are computationally intensive, and rely on crude models of the dynamics. In our recently initiated project, we explore how physics-informed machine learning models, coupled with topological methods for the robust characterization of conformations and their dynamics, will enable the faithful characterization and manipulation of complex self-organising phenomena of nanoparticles.

Speaker: Kaiwen Cai

Title: Uncertainty Estimation for 3D Dense Prediction via Cross-Point Embeddings

Abstract: Dense prediction tasks are common for 3D point clouds, but the uncertainties inherent in massive points and their embeddings have long been ignored. In this work, we present CUE, a novel uncertainty estimation method for dense prediction tasks in 3D point clouds. Inspired by metric learning, the key idea of CUE is to explore cross-point embeddings upon a conventional 3D dense prediction pipeline. Specifically, CUE involves building a probabilistic embedding model and then enforcing metric alignments of massive points in the embedding space. We also propose CUE+, which enhances CUE by explicitly modelling crosspoint dependencies in the covariance matrix. We demonstrate that both CUE and CUE+ are generic and effective for uncertainty estimation in 3D point clouds with two different tasks: (1) in 3D geometric feature learning we for the first time obtain well-calibrated uncertainty, and (2) in semantic segmentation we reduce uncertainty’s Expected Calibration Error of the state-of-the-arts by 16.5%. All uncertainties are estimated without compromising predictive performance.

Speaker: Pablo Lyn Guerrero

Title: TBC

Abstract: TBC