I was lucky to be given the opportunity to present my two recent (ongoing) projects at the Champalimaud Neuro-cybernetics Symposium. The Symposium was organised by Shreya Saxena, Memming Park, and Guillaume Hennequin, and of course the Champalimaud foundation at the Center for the Unknown (more information here).
I presented two posters on my current work: one on Contextual reconfiguration of V1 representations from top-down feedback and one on Discovering three-factor learning rules for training RNNs (or “Meta-learning three-factor learning rules for structured credit assignment in recurrent neural networks” as was the official submitted title). For the abstracts go here or here if the link stops working. Here are also the certificates of the two poster presentations: V1 reconfiguration and meta-learning three-factor rules for RNNs .
I got very excited because I received very useful feedback and ideas on how to improve my gradients for the second project. I pulled an all-nighter to implement the updates into my code and to submit to COSYNE. I am deeply grateful for those who came by and were willing to discuss and offered solutions on the points where I needed help!
These two projects are both solo projects, and those who know me know the reasons, why it is safer to work indepentendly under the current circmstances.
By the way, I find the graphic design of the symposium website fantastic! You can find more of her work >>> here