Speaker: Gavin Gray
Title: Neural Arts and Crafts
Abstract: In just half an hour, learn to transfer artistic styles with deep convolutional neural networks! Continuing from Matt’s talk at the last Journal Club I’m going to be focusing on the Neural Style Transfer paper that recently came out. Bring a laptop and I’ll have everything set up for you to use in the cloud. If you don’t bring a laptop, then bring an idea and we’ll run it somehow. If you don’t have an idea, then bring a beer and watch this setup fall apart on the day.
Speaker: Alina Selega
Title: Deconvolving signals from a heterogeneous population (or Something Actually Relevant to My Work This Time)
Abstract: Disclaimer: What follows next might look like a boring abstract from a “real” paper… That’s because this is my actual PhD work, freshly out of a collaboration so new that I am still in Italy!
So, I can tell you why I think what I’m doing is (could be) important (maybe, I hope?) on an example of a real project. Or, if you’re allergic to words like “gene expression” or “clinical research”, I will throw in some nice photos of hikes I’ve done and maybe even qualcosa di dolce!
In clinical research, the main focus understandably lies on characterising molecular changes in diseased tissues, in order to develop new treatments. However, the analysis is almost always obscured by the heterogeneous tissue composition. For example, pathological cases of neurodegenerative diseases are often accompanied by neuronal death, which could misleadingly lead to measuring reduced neuron-specific expression. I am going to talk about a regression-based method to deconvolve signals from a heterogeneously composed population and also give examples of how measured effects can dramatically differ when accounting for this or not.
Citation: Kuhn, Alexandre, et al. “Population-specific expression analysis (PSEA) reveals molecular changes in diseased brain.” Nature methods 8.11 (2011): 945-947.