Figure 6
The bar graph compares the final error rates for different activation functions on a fully optimized model. The x-axis represents the activation functions, including Sigmoid, Tanh, Relu, Gelu, and Swish. The y-axis represents the error rate, ranging from 0 to 5. The graph shows five vertical bars, each representing a different activation function. The Sigmoid function has the highest error rate, approximately 4.5. The Tanh, Relu, Gelu, and Swish functions have significantly lower error rates, all below 1. The Tanh and Swish functions have the lowest error rates, around 0.5. The Relu and Gelu functions have slightly higher error rates, around 0.75. All values are approximated.Final error rates for different activation functions on the fully optimized model