Towards explainable deep neural networks (xDNN)

被引:161
作者
Angelov, Plamen [1 ]
Soares, Eduardo [1 ]
机构
[1] Univ Lancaster, LIRA Res Ctr, Sch Comp & Commun, Lancaster LA1 4WA, England
关键词
Explainable AI; Interpretability; Prototype-based models; Deep-learning; MODELS;
D O I
10.1016/j.neunet.2020.07.010
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper, we propose an elegant solution that is directly addressing the bottlenecks of the traditional deep learning approaches and offers an explainable internal architecture that can outperform the existing methods, requires very little computational resources (no need for GPUs) and short training times (in the order of seconds). The proposed approach, xDNN is using prototypes. Prototypes are actual training data samples (images), which are local peaks of the empirical data distribution called typicality as well as of the data density. This generative model is identified in a closed form and equates to the pdf but is derived automatically and entirely from the training data with no user- or problem-specific thresholds, parameters or intervention. The proposed xDNN offers a new deep learning architecture that combines reasoning and learning in a synergy. It is non-iterative and non-parametric, which explains its efficiency in terms of time and computational resources. From the user perspective, the proposed approach is clearly understandable to human users. We tested it on challenging problems as the classification of different lighting conditions for driving scenes (iROADS), object detection (Caltech-256, and Caltech-101), and SARS-CoV-2 identification via computed tomography scan (COVID CT-scans dataset). xDNN outperforms the other methods including deep learning in terms of accuracy, time to train and offers an explainable classifier. (C) 2020 Elsevier Ltd. All rights reserved.
引用
收藏
页码:185 / 194
页数:10
相关论文
共 44 条
[1]  
Angelov P., 2012, AUTONOMOUS LEARNING
[2]  
Angelov P., 2019, Empirical approach to machine learning
[3]   Deep rule-based classifier with human-level performance and characteristics [J].
Angelov, Plamen P. ;
Gu, Xiaowei .
INFORMATION SCIENCES, 2018, 463 :196-213
[4]   A Generalized Methodology for Data Analysis [J].
Angelov, Plamen P. ;
Gu, Xiaowei ;
Principe, Jose C. .
IEEE TRANSACTIONS ON CYBERNETICS, 2018, 48 (10) :2981-2993
[5]   Prototype-based models in machine learning [J].
Biehl, Michael ;
Hammer, Barbara ;
Villmann, Thomas .
WILEY INTERDISCIPLINARY REVIEWS-COGNITIVE SCIENCE, 2016, 7 (02) :92-111
[6]  
Biehl Michael., 2013, International Workshop on Brain-Inspired Computing, P100
[7]  
Bishop CM., 2006, Springer Google Schola, V2, P1122, DOI [10.5555/1162264, DOI 10.18637/JSS.V017.B05]
[8]   Improved support vector machine classification algorithm based on adaptive feature weight updating in the Hadoop cluster environment [J].
Cao, Jianfang ;
Wang, Min ;
Li, Yanfei ;
Zhang, Qi .
PLOS ONE, 2019, 14 (04)
[9]   The devil is in the details: an evaluation of recent feature encoding methods [J].
Chatfield, Ken ;
Lempitsky, Victor ;
Vedaldi, Andrea ;
Zisserman, Andrew .
PROCEEDINGS OF THE BRITISH MACHINE VISION CONFERENCE 2011, 2011,
[10]  
Doshi-Velez F., 2017, stat, V1702