[eigenvalue]

Relu is literally just a linear function that gets clamped to zero at some point, so yes, it's much less computationally intensive than anything involving an exponential function. But I doubt you would get competitive results using such a simple activation.