Visualizing the Function Computed by a Feedforward Neural Network

A method for visualizing the function computed by a feedforward neural network is presented. It is most suitable for models with continuous inputs and a small number of outputs, where the output function is reasonably smooth, as in regression and probabilistic classification tasks. The visualization makes readily apparent the effects of each input and the way in which the functions deviate from a linear function. The visualization can also assist in identifying interactions in the fitted model. The method uses only the input-output relationship and thus can be applied to any predictive statistical model, including bagged and committee models, which are otherwise difficult to interpret. The visualization method is demonstrated on a neural network model of how the risk of lung cancer is affected by smoking and drinking.

[1]  David Sharp.  From "black box" to bedside, one day , 1995, The Lancet.

[2]  Jeffrey L. Elman,et al.  Analyzing Cross-Connected Networks , 1993, NIPS.

[3]  J. Shavlik,et al.  Extracting Reened Rules from Knowledge-based Neural Networks Keywords: Theory Reenement Integrated Learning Representational Shift Rule Extraction from Neural Networks , 1992 .

[4]  Dennis Sanger,et al.  Contribution analysis: a technique for assigning responsibilities to hidden units in connectionist networks , 1991 .

[5]  R. Tibshirani,et al.  Generalized Additive Models , 1991 .

[6]  Thomas R. Shultz,et al.  Analysis of Unstandardized Contributions in Cross Connected Networks , 1994, NIPS.

[7]  Tony Plate,et al.  A Comparison between Neural Networks and other Statistical Techniques for Modeling the Relationship between Tobacco and Alcohol and Cancer , 1996, NIPS.

[8]  Halbert White,et al.  Bootstrapping Confidence Intervals for Clinical Input Variable Effects in a Network Trained to Identify the Presence of Acute Myocardial Infarction , 1995, Neural Computation.

[9]  L. Moseholm,et al.  Pulmonary function changes in asthmatics associated with low‐level SO2 and NO2 air pollution, weather, and medicine intake , 1993, Allergy.

[10]  Heekuck Oh,et al.  Neural Networks for Pattern Recognition , 1993, Adv. Comput..

[11]  Reinhard Blasig.  GDS: Gradient Descent Generation of Symbolic Classification Rules , 1993, NIPS.

[12]  Jeremy Wyatt,et al.  Nervous about artificial neural networks? , 1995, The Lancet.

[13]  R. Nakano,et al.  Medical diagnostic expert system based on PDP model , 1988, IEEE 1988 International Conference on Neural Networks.

[14]  David J. C. MacKay,et al.  The Evidence Framework Applied to Classification Networks , 1992, Neural Computation.

[15]  W. Baxt.  Analysis of the clinical variables driving decision in an artificial neural network trained to identify the presence of myocardial infarction. , 1992, Annals of emergency medicine.

[16]  T. Plate.  ACCURACY VERSUS INTERPRETABILITY IN FLEXIBLE MODELING : IMPLEMENTING A TRADEOFF USING GAUSSIAN PROCESS MODELS , 1999 .

[17]  Jenq-Neng Hwang,et al.  Regression modeling in back-propagation and projection pursuit learning , 1994, IEEE Trans. Neural Networks.