search

Explainable AI by Combining Statistical and Logical Methods

Machine learning is prevalent in big data analytics. At present, almost all machine learning system operates statistically or optimize blindly, regarded by academia and industry. The causal logic behind the results remains a black box, limiting the application of machine learning. Starting from a structure approach to causal reasoning, our research uses machine learning to improve application accuracy, and rules to ensure the interpretability of results. At the same time, a new rule system combined with a machine learning model has been put forward. It can support plug-and-play and deal with the potential semantic relationships in data.

Research Areas

Beginning with building the logical relationship between machine learning’s input and output flow on the graph data structure, then we put an effort to effectively reveal the reasoning logic of machine learning in natural language processing, intelligent question answering, semantic model analysis, etc., based on the detailed topological structure and association relationships of graph data levels, attributes, and categories, etc. It enables us with targeted optimization, further improves machine learning performance, and expands the application range. We also research a new type of rule system that combines logic and machine learning models, maintaining the logical reasoning relationship between data, while use machine learning models to enhance semantic expression capabilities, ultimately achieve a real unity of models and logic. This system is prevalent in fields such as data quality and association analysis.

Related Publications

  • Capturing Associations in Graphs

    Authors: Wenfei Fan, Ruochun Jin, Muyang Liu, Ping Lu, Chao Tian, Jingren Zhou Published in: International Conference on Very Large Data Bases (VLDB 2020), A...

    Wenfei Fan, Ruochun Jin, Muyang Liu, Ping Lu, Chao Tian, Jingren Zhou International Conference on Very Large Data Bases (VLDB 2020) International Conference on Very Large Data Bases (VLDB 2020)
  • Monotonic Cardinality Estimation of Similarity Selection: A Deep Learning Approach

    Authors: Yaoshu Wang, Chuan Xiao, Jianbin Qin, Xin Cao, Yifang Sun, Wei Wang, Makoto Onizuka Name of Conference: ACM Conference on Management of Data(SIGMOD...

    Yaoshu Wang, Chuan Xiao, Jianbin Qin, Xin Cao, Yifang Sun, Wei Wang, Makoto Onizuka ACM Conference on Management of Data(SIGMOD 2020) ACM Conference on Management of Data (SIGMOD 2020)
show more