جعبه سیاه در مقابل جعبه سفید
Abstract
I. Introduction
II. Black-Box Approach
III. White-Box Approach
IV. Fusing Black- and White-Box Approaches
V. Measuring the Effectiveness of the Applied Model
Authors
Figures
References
Abstract
Nowadays, in the international scientific community of machine learning, there exists an enormous discussion about the use of black-box models or explainable models; especially in practical problems. On the one hand, a part of the community defends that black-box models are more accurate than explainable models in some contexts, like image preprocessing. On the other hand, there exist another part of the community alleging that explainable models are better than black-box models because they can obtain comparable results and also they can explain these results in a language close to a human expert by using patterns. In this paper, advantages and weaknesses for each approach are shown; taking into account a stateof-the-art review for both approaches, their practical applications, trends, and future challenges. This paper shows that both approaches are suitable for solving practical problems, but experts in machine learning need to understand the input data, the problem to solve, and the best way for showing the output data before applying a machine learning model. Also, we propose some ideas for fusing both, explainable and blackbox, approaches to provide better solutions to experts in real-world domains. Additionally, we show one way to measure the effectiveness of the applied machine learning model by using expert opinions jointly with statistical methods. Throughout this paper, we show the impact of using explainable and black-box models on the security and medical applications.
Introduction
Three decades ago, a part of the international scientific community working on computer science was focused on creating machine learning models for solving theoretical challenges [1], [2]. Nowadays, there exist both theoretical and practical progress in computer science. However, the international scientific community working on machine learning has begun an important debate about the relevance of the blackbox approach and the explainable approach (a.k.a white-box approach) from a practical point of view [3], [4]. On the one hand, the international scientific community of machine learning has labeled as black-box models all those proposals containing a complex mathematical function (like support-vector machine and neuronal networks) and all those needing a deep understanding of the distance function and the representation space (like k-nearest neighbors), which are very hard to explain and to be understood by experts in practical applications [4]–[6]. On the other hand, those models based on patterns, rules, or decision trees are labeled as white-box models; and they, usually, ca be understood by experts in practical applications due to they provide a model closer to the human language [7]–[9]. There has been a trend of moving away from blackbox models towards white-box models, particularly for critical industries such as healthcare, finances, and military (e.g. battlefields).