Involution是由Li等人于2021年提出的新型神经网络算子。该算子基于卷积(Convolution)设计而来，具有与卷积相反的基本特征，故得名，也常被称作内卷、合卷。本文统一称为 Involution。
本文记录笔者在本科毕业设计中对 Involution 的学习与实现过程，包括基础原理梳理、结构改造思路，以及并行化尝试与性能相关思考。
Involution is a neural network operator proposed by Li et al. in 2021. It is designed with characteristics opposite to standard convolution, and is often referred to as an inverse-style convolutional operator.
This post records my study and implementation notes from undergraduate thesis work, including core ideas, structural modifications, and early parallelization attempts for acceleration.
