It can be clearly observed that there is an obvious trend in the field of data intelligence. The complexity, diversity and synergy of computing are becoming more and more obvious. The complexity is mainly reflected in the computing model, including the most popular ones like deep learning, reinforcement learning, migration Learning such mainstream machine learning methods involves tens of millions of parameters. How to ensure its interpretability and credibility has become a core risk issue. This is also the original intention of launching Trust Day. The brains of experts in the industry discuss together and brainstorm together.
Data and intelligence are two sides of the same coin. Data is an important factor driving computational intelligence. As models become more complex, more data needs to be used, ranging from traditional numerical data to text data to multimedia data, including image data and video data. , and today’s network data, in the case of large-scale fusion of data and collaborative intelligence, how to design better algorithms to evaluate the contribution of each data in the model, some new academic directions have also been proposed, such as data auditing ( Data Auditing), data auditing is different from general financial auditing. When there are huge multi-source and heterogeneous data in the model, as a market participant, it is more necessary to measure the contribution of data in the model in a rigorous and quantitative way, so as to provide Participating parties distribute responsibilities and interests, which is also an important part of data intelligence trust.
Reference: https://mp.weixin.qq.com/s/yokroFowIe3AIKgpNU_Igg
