Skip to main content
. 2024 Oct 8;13:283. doi: 10.1038/s41377-024-01651-7

Fig. 4. Experimental results for fashion-MNIST binary image classification.

Fig. 4

a Information flow for the PNCA trained to classify images of sneakers and trousers, beginning with alive cell masking, followed by t = 21 iterations of the trained PNCA. The predicted image label is obtained by global average pooling and softmax classification of the final self-organized alive cells. Confusion matrices for (b) idealized simulation model, (c) noisy simulation model, and (d) experiment