Neural network architecture and rodents brain

We see twice - or why the fly’s brain is better than our big brains when swapping is involved.

Vision came early in evolution; trilobites could see more than 500 million years ago. Yet, our understanding of visual processes is cortex-centric, a recent structure. Similarly, deep-neural network architectures are often loosely based on the visual cortex. One might ask: does a computerk really need to “consciously” see? In this presentation, I will share completely novel findings from brain implant data in rodents showing that there exists an evolutionary ancient brain region, the superior colliculus, that we share with mosquitos, fishes and birds, which can compute complex scenes and can even process categorical information. We will then discuss how we could use this fast acting, wide-reaching circuitry to come up with novel neural network architectures. These new networks could be useful for fast-acting machine-vision models with far reaching potential, such as for scene navigation.

Agenda:

3:45pm - 4:00pm: Arrival & Networking
4:00pm - 4:15pm: News & Intro
4:15pm - 5:15pm: Neural network architecture and rodents brain
5:15pm - 5:30pm: Virtual Snack & Networking

About the main speaker:

William John Redmond is a computational neuroscientist bridging machine-vision and visual neurosciences. He has received his PhD in Advanced Medicine at Macquarie University, where he worked on cannabinoid receptor signaling pathway. He then completed 3 postdoctoral positions in Montréal, Leuven and Sydney where he focused on system neuroscience, brain-machine interfaces and optogenetic.

Logistics

This is an online event. Participants will receive the link to the video conference after they register.
The schedule above is using Eastern Daylight Time (GMT-4).