Making machine learning models interpretable

摘要: Data of different levels complexity and ever growing diversity characteristics are the raw materials that machine learning practitioners try to model using their wide palette methods tools. The obtained models meant be a synthetic representation available, observed data captures some intrinsic regularities or patterns. Therefore, use techniques for analysis can understood as problem pattern recognition or, more informally, knowledge discovery mining. There exists gap, though, between modeling extraction. Models, de- pending on employed, described in diverse ways but, order consider has been achieved from description, we must take into account human cog- nitive factor any extraction process entails. These such rendered powerless unless they interpreted ,a nd interpretation follows rules go well beyond techni- cal prowess. For this reason, interpretability is paramount quality should aim achieve if applied practice. This paper brief introduction special session interpretable learning, organized part 20 th European Symposium Artificial Neural Networks, Computational In- telligence Machine Learning. It includes discussion several works accepted session, with an overview context wider research models.