weights that connect neurons can only be multiplied. now i am looking at activation functions like `exponent`. i just doubt a neural network regression would be

weights that connect neurons can only be multiplied. now i am looking at activation functions like `exponent`. i just doubt a neural network regression would be able to predict rlly high values accurately for something like E = mc^2.