Figure 1 | Scientific Reports

Figure 1

From: Run-off election-based decision method for the training and inference process in an artificial neural network

Figure 1

(a) Constituents of FCSN for the 28 × 28 shoe image inference process. Each input pixel is linked to the output neurons consisting of ‘Sandal’, ‘Sneaker’ and ‘Boot’ with a synaptic weight. (b) Configuration of the activation function showing 3 output results for each input image (between − 1 and 1) for the ideal cases (upper and middle line) and worst case (lower line). (c) 18,000 training images and (d) 3,000 test images for the shoe data set. (e) Flow chart of one epoch of the training/test algorithm, where the yellow box shows that the filter evaluation is only executed if the difference between the first and second activation function values is smaller than the predetermined δ-value.

Back to article page