From Neurons to Networks: Core Principles Linking Biological and Artificial Intelligence
How AI activation functions and weights map to firing rates in biological neurons.
Circuit level computation: excitatory-inhibitory networks, feedback loops, and layered architecture paralleling AI computational graphs
Brain-to-ML connections: supervised and reinforcement learning via visual pathways, dopamine rewards, state-action representations, and habit formation
Nonlinear and temporal dynamics across biological scales
Diplomacy game case study demonstrating how strategic reasoning emerges in brains and AI through self-play, imitation learning, and reinforcement learning
To Attend Online:
Add to calendar: https://shorturl.at/nhyPW
Gmeet link: meet.google.com/cad-aeyt-vkp
About the speaker:
Dr Selva Maran is the Co-founder/CTO for the startup that focuses on early diagnosis and prevention of Dementia.
He completed his medical degree at Madras Medical College. Later he did PhD in Computational Neuroscience, at Louisiana State University Health Sciences Centre (LSUHSC) in New Orleans, USA and a postdoc at Emory University in Atlanta, USA.
His PhD and postdoctoral research areas were focused on understanding information processing by building computational models to simulate spike train patterns observed in biological neurons
Over the past seven years, his work has been focused on Artificial Intelligence applications in healthcare, especially leveraging large scale multimodal datasets to improve Early Diagnosis, Precision Diagnosis and Clinical Care
Speaker LinkedIn: HERE
Pre Read
Deep Reinforcement Learning and its Neuroscientific Implications: https://arxiv.org/abs/2007.03750
Reinforcement Learning, Fast and Slow: https://doi.org/10.1016/j.tics.2019.02.006
CCN course: https://www.youtube.com/@RandallOReilly/courses , https://compcogneuro.org/
Modelling Strong and Human-Like Gameplay with KL-Regularized Search: https://arxiv.org/abs/2112.07544
Mastering the Game of No-Press Diplomacy via Human-Regularized Reinforcement Learning and Planning: https://arxiv.org/abs/2210.05492