ANC Workshop - Oisin Mac Aodha, Yuelin Yao

Tuesday, 14th March 2023

Learning from Visual Data in the Wild - Oisin Mac Aodha

Abstract: The wealth and complexity of visual information potentially observable by artificial systems deployed in the real world vastly exceeds the comparative simplicity of our carefully curated benchmark computer vision datasets. To operate safely and reliably in challenging environments, these systems need to be able to correctly recognize previously learned concepts, not confuse these known concepts with novel ones, and be able to differentiate novel concepts so that they can be grouped and efficiently learned. In this talk I will discuss some recent work from my group on problems related to learning from vision data “in the wild”. Specifically, I will cover recent work on self-supervised learning of shape from images, incremental category discovery, and how to encode priors about where categories are likely to be observed.

Stator: Define cell identities in scRNA-seq by higher-order interactions - Yuelin Yao

Abstract: Advances in scRNA-seq techniques have led to successes in quantifying gene expression at a single-cell level, and thus make it possible to uncover novel cell types/subtypes or states among complex cell populations. However, traditional clustering analysis is not able to reveal continuous spectrum of cell states. We introduce Stator, a model-independent estimation to identify higher-order interactions among genes and define cell states. We use an iterative MCMC algorithm to infer a quasi-causal graph of conditional dependencies among the genes and estimate the model-free interactions (MFIs). Hierarchical clustering of significantly deviating states of MFIs by cell assignment revealed substructures in the cell population corresponding to diverse cell states. We demonstrate the ability of Stator to extract biologically meaningful cell states using publicly available scRNA-seq datasets including hepatocellular carcinoma dataset, developmental mouse brain dataset and intestine dataset.

Event type: Workshop

Date: Tuesday, 14th March 2023

Time: 11:00

Location: G.03

Speaker(s): Oisin Mac Aodha, Yuelin Yao

Chair/Host: Hancong Wu