Explainable AI: A Review of Machine Learning Interpretability Methods

Abstract

Recent advances in artificial intelligence (AI) have led to its widespread industrial adoption, with machine learning systems demonstrating superhuman performance in a significant number of tasks. However, this surge in performance, has often been achieved through increased model complexity, turning such systems into “black box” approaches and causing uncertainty regarding the way they operate and, ultimately, the way that they come to decisions. This ambiguity has made it problematic for machine learning systems to be adopted in sensitive yet critical domains, where their value could be immense, such as healthcare. As a result, scientific interest in the field of Explainable Artificial Intelligence (XAI), a field that is concerned with the development of new methods that explain and interpret machine learning models, has been tremendously reignited over recent years. This study focuses on machine learning interpretability methods; more specifically, a literature review and taxonomy of these methods are presented, as well as links to their programming implementations, in the hope that this survey would serve as a reference point for both theorists and practitioners.

Anthropogenic influence on extreme precipitation over global land areas seen in multiple observational datasets

“…However, even with adequate sampling of natural variability in the training dataset, the underestimation of the precipitation response to natural forcings such as volcanic activities and natural variability such as El Nino Southern Oscillation in GCMs could still affect the results 62 . We also note that different ANN visualization techniques are available [63][64][65] , and those should be explored to understand the sensitivity of the extracted fingerprints to the ANN visualization technique. Despite these limitations, it is clear that ANN DAI methods with ANN visualization techniques are very useful and efficient in identifying the human influence on variables that are highly uncertain in GCMs, and poorly characterized in observations, such as extreme precipitation.…”

Deep Learning and Explainable Artificial Intelligence Techniques Applied for Detecting Money Laundering–A Critical Review

“…Weber et al [97] presented a comprehensive review of machine learning interpretability methods using 4 categories -(i) Methods for explaining black box models (17 methods focused on DL method interpretations, 16 methods that can explain any black-box model), (ii) methods for creating white-box models (5 methods), (iii) methods that promote fairness and restrict discrimination, (iv) methods that analyses the sensitivity of model predictions (28 methods). Most interpretability methods are focused on DL, largely ruled by neural networks, and are experimented with image classification explanation.…”

Machine learning aided noise filtration and signal classification for CREDO experiment

“…First, it provides a lower bound of the classification accuracy for the more elaborate, eg. neural network based classifiers, while being entirely explainable [15]. Second, relying on the simple heuristics they are much less computationally challenging, so can be effortlessly implemented as online triggers working with limited resources of smartphones.…”