机器学习的知识产权保护途径初探

A Preliminary Study on the Protection of Intellectual Property Rights in Machine Learning

进行机器学习应用开发、运营的公司往往投入大量人力物力,但不少公司常怀觊觎之心,意图通过盗用、套改机器学习模型、训练集、源代码等方式,以达到节约成本、快速赚取利润的目的。公司如何进行机器学习系统的知识产权保护,避免相关成果被竞争对手轻易获取?具体的保护方式有哪些?如何选择最有利于自己的维权途径?怎样更好举证并获得快速审理、高额赔偿?本文对机器学习相关的著作权、数据库、专利、商业秘密、不正当竞争等保护途径进行了梳理,并对公司在选择保护途径中应注意的问题进行了提示。

Companies that develop machine learning applications often invest a lot of manpower and material resources,but many competitors often covet it,intending to save costs and earn quick money by embezzling or altering machine learning models,training sets,source codes,etc.How do companies protect the intellectual property rights of machine learning systems to prevent related achievements from being easily picked by competitors?What are the specific protection methods?How to choose the most favorable way to protect their rights?How to provide evidence and obtain fast trial and high compensation?This article sorts out the protection methods related to machine learning,such as copyright,database,patent,trade secrets,unfair competition,etc.,and suggests the problems that companies should pay attention to in choosing protection methods.