Skip to yearly menu bar Skip to main content

Invited Talk

Engineering Principles From Stable and Developing Brains

Saket Navlakha

Area 1 + 2


Robust, efficient, and low-cost networks are advantageous in both biological and engineered systems. First, I will describe a joint computational-experimental approach to explore how neural networks in the brain form during development. I will discuss how the brain uses a very uncommon and surprising strategy to build networks and how this idea can be used to enhance the design and function of energy-efficient distributed networks. Second, I will describe how two fundamental plasticity rules (LTP and LTD) help neural networks approach desirable synaptic weight distributions in a gradient-descent-like manner. I will derive connections between different experimentally-derived forms of these rules and distributed algorithms commonly used to regulate traffic flow on the Internet. Our work is motivated by the study of “algorithms in nature”.

Chat is not available.