Optimizing Genetically-Driven Synaptogenesis
CoRR(2024)
摘要
In this paper we introduce SynaptoGen, a novel framework that aims to bridge
the gap between genetic manipulations and neuronal network behavior by
simulating synaptogenesis and guiding the development of neuronal networks
capable of solving predetermined computational tasks. Drawing inspiration from
recent advancements in the field, we propose SynaptoGen as a bio-plausible
approach to modeling synaptogenesis through differentiable functions. To
validate SynaptoGen, we conduct a preliminary experiment using reinforcement
learning as a benchmark learning framework, demonstrating its effectiveness in
generating neuronal networks capable of solving the OpenAI Gym's Cart Pole
task, compared to carefully designed baselines. The results highlight the
potential of SynaptoGen to inspire further advancements in neuroscience and
computational modeling, while also acknowledging the need for incorporating
more realistic genetic rules and synaptic conductances in future research.
Overall, SynaptoGen represents a promising avenue for exploring the
intersection of genetics, neuroscience, and artificial intelligence.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要