Are artificial neural networks black boxes?

By José Manuel Benítez, J. Castro, I. Requena. 1997

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Explaining Explanations: An Overview of Interpretability of Machine Learning (2018, background): “There has also been work on transforming neural network to fuzzy rules [30], by transforming each neuron into an approximate rule.”
Neural networks for classification: a survey (2000, background): “Numerous insights have also been gained into the neural networks in performing classification as well as other tasks [23], [169].”
Avoiding Pitfalls in Neural Network Research (2007, unknown): “[10] and Castro et al.”
Deep learning classifiers for hyperspectral imaging: A review (2019, background): “The “black box” nature of the training procedure is also a disadvantage, being the model’s internal dynamics very hard to interpret (Benítez et al., 1997; Lipton, 2016).”
The impact of modelling choices in the predictive performance of richness maps derived from species‐distribution models: guidelines to build better diversity models (2013, unknown): “2009; Bisrat et al. 2012). Such good results indicate that random forest is a promising algorithm to generate SSDMs, but taking into account that the algorithm results are prone to overfitting, and show poor performance when transferred across time or space (Heikkinen, Marmion & Luoto 2012). The sec”
Explaining Explanations: An Approach to Evaluating Interpretability of Machine Learning (2018, background): “There has also been work on transforming neural network to fuzzy rules [30], by transforming each neuron into an approximate rule.”
DeepRED - Rule Extraction from Deep Neural Networks (2016, methodology): “Examples for decompositional approaches are the KT method that heuristically searches for input combinations that let a neuron fire [7], a more efficient implementation of the latter [20], and an algorithm that transforms NNs to fuzzy rules [3].”
Calibration of Pre-trained Transformers (2020, background): “Neural networks have seen wide adoption but are frequently criticized for being black boxes, offering little insight as to why predictions are made (Benı́tez et al., 1997; Dayhoff and DeLeo, 2001; Castelvecchi, 2016) and making it difficult to diagnose errors at test-time.”
Extracting symbolic rules from trained neural network ensembles (2003, background): “[3] establish the equality between a certain class of neural networks and fuzzy rule based systems so that the network could be explained by a set of fuzzy rules.”