Figure 3
Visualization of the deep neural network architecture of the TensorFlow model used in the present analysis with (A) the full feature set (d = 11 pain-threshold related variables), d = 4 variables that had resulted from the feature selection steps shown in Fig. 4 as “reduced” feature set (B) (darker blue columns Fig. 4A), or d = 2 variables that had resulted from further narrowing the feature set to the “sparse” feature set (C) (darkest blue columns Fig. 4A). Variables X1..11 are the input variables, H1..n denotes the hidden layers, and the binary output layer is shown on the right. In addition, the respective activation functions are given on the top of each layer. The graph shows the neural networks architectures as implemented with “TensorFlow” and run on an NVIDIA GeForce RTX 3060 GPU (NVIDIA Corporation, Santa Clara, CA, USA). The figure was created using Python version 3.8.13 for Linux (https://www.python.org), with the seaborn statistical data visualization package (https://seaborn.pydata.org40) and Python code modified from https://towardsdatascience.com/deep-learning-with-python-neural-networks-complete-tutorial-6b53c0b06af0.
