If I change the activation function of VGG16, will the accuracy increase?

If I change the activation function of VGG16 to switch, ReLU6, LeakyReLU, Mesh, etc., will the accuracy improve? Or does it go down?

[Walter Roberson]

I think you would need to be clear about what kind of data you would be using this for. A change that improves one kind of input might do worse on a different kind of input m

[ys]

For example CIFAR-10 or CIFAR-100

Answer 1

[Sanyam]

This cannot be directly said, as kind of activation function to use is also a hyperparameter. What you can do is to check the performance of your model with different activation functions using cross-validation or using a seperate validation dataset. Then pick up the activation function which gives the best performance

Hope that helps. Thanks!!