实验性的神经网络块级去相关层的实现，用于比较与批量归一化和普通架构的性能差异

【实验性的神经网络块级去相关层的实现，用于比较与批量归一化和普通架构的性能差异】' Block-wise Decorrelation Layer Experiments - This repository contains experiments with block-wise decorrelation layers in neural networks, comparing them with batch normalization and plain architectures.' GitHub: github.com/cloneofsimo/blocked-decorr 神经网络 去相关层 机器学习 效率优化