Credit assignment through critical period plasticity

Dr. David Ehrlich is an assistant professor of Integrative Biology at the University of Wisconsin-Madison.

Across the brain, learning relies on plasticity at specific synapses guided by instructive signals. Yet behavior emerges from distributed neural circuits, requiring localized, instructed plasticity to be translated into behaviorally relevant changes across complete circuits. No general mechanism is known for enforcing logical organization of circuits such that local plasticity supports circuit-wide performance. I'll share our surprising recent discovery that instructive signals themselves can achieve logical network organization through a simple Hebbian development rule. Leveraging a dedicated instructive pathway in zebrafish, we show instructive climbing fibers coordinate activity-dependent axon remodeling downstream of their direct synaptic targets, leading to lasting changes in circuit function. We then prove mathematically that networks shaped by such 'instructed maturation' become logically organized for subsequent learning. We use simulations to show that instructed maturation accelerates future learning and sculpts networks that outperform random networks using machine learning algorithms in idealized conditions. This long-term advantage comes at substantial cost, because maturation scrambles memory expression by remodeling circuits downstream of learning sites, suggesting instructed maturation should be transient. Consistently, developmental remodeling in zebrafish was confined to a brief critical period days after fish started swimming, interpreted as a window to use experience for instructed maturation and then abruptly halt it for stable learning. Together, these results reveal a second function for instructive signals outside of learning, establishing circuit organization that optimally aligns local plasticity with downstream function for rapid and effective circuit-based learning.

Learn more

Hosted by Dr. Mike Mauk

To subscribe to our Seminars Series mailing list, please visit https://utlists.utexas.edu/sympa/info/neu-seminars and click on "Subscribe" in the left-hand column.  

Please contact Neuro.Seminar.Series@austin.utexas.edu with any questions.