Campus Only Senior Thesis
Bachelor of Arts
Professor Sarah Marzen
Professor John Milton
2020 Rachel Taubman
Biological neural networks encode predictive information about their environment with high energy efficiency and minimal learning supervision. In this study, a Hopfield-like recurrent neural network with two biologically-based learning rules consistently improves both prediction and energy efficiency in multiple parameter regimes with a two-state hidden Markov Model stimulus. The network also exhibits emergent synaptic normalization, which suggests that this feature observed in neurons may emerge from an interaction of other learning rules.
Taubman, Rachel, "Recurrent Neural Network with Homeostatic Plasticity Learning Rules Improves Prediction and Energy Efficiency and Leads to Emergent Synaptic Normalization" (2020). Scripps Senior Theses. 1590.
This thesis is restricted to the Claremont Colleges current faculty, students, and staff.