Interpreting neural-network connection weights

By G. Garson. 1991

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 nonlinear classification decisions with deep Taylor decomposition (2015, methodology): “The w(2)-rule resembles the rule by [46, 44] for determining the importance of input variables in neural networks, where absolute values of wij are used in place of squared values.”
Machine Learning Methods Without Tears: A Primer for Ecologists (2008, methodology): “Other approaches include Garson’s algorithm (Garson 1991), partial derivatives (Dimopoulos et al. 1995), a sensitivity analysis to determine the spectrum of input variable contributions in the neural network (Lek et al. 1996), and a number of pruning algorithms (Bishop 1995), including a…”
Multivariate adaptive regression splines and neural network models for prediction of pile drivability (2016, methodology): “For the BPNN, this is commonly carried out using the method by Garson (1991) and discussed by Das and Basudhar (2006).”
Applications of artificial neural networks for patterning and predicting aquatic insect species richness in running waters (2003, methodology): “Among these algorithms, we can classify in several categories using: (i) the connection weights (Garson 1991; Goh 1995), (ii) the connection weights and a fictitious input matrix considering a successive variation of one input variable while the others are kept constant at a fixed value (Lek et…”
Hybrid neural systems: from simple coupling to fully integrated neural networks (1999, background): “The information learned by a neural network is held in the weights and biases as a result of exposure to the training set [31, 27].”
Review and comparison of methods to study the contribution of variables in artificial neural network models (2003, methodology): “The procedure for partitioning the connection weights to determine the relative importance of the various inputs was proposed first by Garson (1991) and repeated by Goh (1995), see Appendix A. The method essentially involves partitioning the hidden-output connection weights of each hidden neuron i”
A Bayesian regularized artificial neural network for stock market forecasting (2013, background): “This link is stored as a weighted value which provides a measure of the connection between two nodes (Garson, 1991; Goh, 1995).”
An accurate comparison of methods for quantifying variable importance in artificial neural networks using simulated data (2004, methodology): “Illustration of the critical difference between the Connection Weight Approach (Olden and Jackson, 2002b) and Garson's Algorithm (Garson, 1991) that results in their differential ability to correctly identify variable importance in neural networks.”