📄 中文摘要
神经网络长期以来致力于模拟人脑的学习能力。尽管深度神经网络（DNN）在神经元设计上受到大脑的启发，但其训练方法却与生物基础相悖。反向传播作为DNN的主要训练方法，需要大量的计算资源和完全标注的数据集，这在开发和应用中形成了重大瓶颈。通过回归生物仿生，特别是模仿大脑通过修剪学习的方式，可以在不使用标签的情况下，利用数量级更少的计算资源解决各种经典机器学习问题。实验成功地实现了多个语音识别和个性化应用的效果。
细致修剪：一种生物启发的机器学习模型个性化算法
📄 English Summary
Fine-Pruning: A Biologically Inspired Algorithm for Personalization of Machine Learning Models
This research presents a novel algorithm inspired by biological processes, specifically focusing on the pruning mechanisms of the brain to enhance the personalization of machine learning models. Traditional deep neural networks (DNNs) rely heavily on backpropagation, which demands extensive computational resources and fully labeled datasets, creating significant challenges in practical applications. By mimicking the brain's learning process through fine-pruning, this approach demonstrates the capability to address various classical machine learning problems with substantially reduced computational costs and without the need for labeled data. Experimental results indicate successful personalization in multiple speech recognition applications, showcasing the potential of this biomimetic strategy.