Figure 2.

A network geometry diagram illustrates the outcomes of accuracy of the network meta-analysis involving direct comparisons of different algorithms. The thickness of the lines corresponds to the number of direct comparisons between each pair of interventions (LR, logistic regression; MP, multilayer perceptron; MCI, mild cognitive impairment; SVM, simple vector machine; LDS, low density separation; ELM, extreme learning machine; FFR, feed forward network; CNN, convolutional neural network).