Complexity of Neural Network Training and ETR: Extensions with Effectively Continuous Functions
This work addresses foundational complexity questions in machine learning for researchers, establishing theoretical limits on algorithmic solvability for neural network training with specific activation functions.
The paper investigates the computational complexity of training neural networks with various activation functions, showing that sigmoid activation leads to equivalence with the existential theory of the reals with exponential functions (open decidability), while sinusoidal activation results in undecidability, and provides general upper bounds in the arithmetical hierarchy.
We study the complexity of the problem of training neural networks defined via various activation functions. The training problem is known to be existsR-complete with respect to linear activation functions and the ReLU activation function. We consider the complexity of the problem with respect to the sigmoid activation function and other effectively continuous functions. We show that these training problems are polynomial-time many-one bireducible to the existential theory of the reals extended with the corresponding activation functions. In particular, we establish that the sigmoid activation function leads to the existential theory of the reals with the exponential function. It is thus open, and equivalent with the decidability of the existential theory of the reals with the exponential function, whether training neural networks using the sigmoid activation function is algorithmically solvable. In contrast, we obtain that the training problem is undecidable if sinusoidal activation functions are considered. Finally, we obtain general upper bounds for the complexity of the training problem in the form of low levels of the arithmetical hierarchy.