generateActivationFunctionWithAccuracy

Generates C code for activation functions with exact numerical consistency. Matches existing DefaultCpuOps implementations for consistency.

Enhanced for numerical accuracy by:

• Using the same mathematical functions as DefaultCpuOps

• Implementing consistent handling of edge cases (NaN, infinity)

• Ensuring exact transcendental function behavior

• Adding input validation for numerical stability

Supported activations:

• ReLU: max(0, x) using fmaxf() - matches DefaultCpuOps.relu()

• Sigmoid: 1 / (1 + exp(-x)) using expf() - matches DefaultCpuOps.sigmoid()

• Tanh: tanh(x) using tanhf() - matches DefaultCpuOps.tanh()

Parameters