Update: I plotted this equation with the left hand side as another variable, and it shoots off into infinity long before the left hand side reaches 2.3. Is this what I stands for, infinity?

Hi epostma,

Yes I do know about the exp(...) in my equation is it meant to be the base of the natural logarithm and not another variable.

As to what I want to do, I'm in my honours year of computer science, and my topic revolves around neural networks. The equation above is just a simplified version of a backpropagation neural network with 3 inputs, 2 hidden nodes and 1 output node using the sigmoid function. I've removed what I can such as the transformation for the output node as that is easy. I basically want to find an inverse function to the one above where y = f(i_1, i_2, ..., i_n) for i_n = f^-1(y, i_1, i_2, ..., i_(n-1)). In the equation above, i_n is just i and all other i's have been condensed into a and b. It would be fantastic if I can work out the inverse function, my thesis is about different data transformations (for which I require the inverse) and it would be extremely useful if I could count a neural network model as a data transformation.

Thanks,

Chewyraver

Hi epostma,

Yes I do know about the exp(...) in my equation is it meant to be the base of the natural logarithm and not another variable.

As to what I want to do, I'm in my honours year of computer science, and my topic revolves around neural networks. The equation above is just a simplified version of a backpropagation neural network with 3 inputs, 2 hidden nodes and 1 output node using the sigmoid function. I've removed what I can such as the transformation for the output node as that is easy. I basically want to find an inverse function to the one above where y = f(i_1, i_2, ..., i_n) for i_n = f^-1(y, i_1, i_2, ..., i_(n-1)). In the equation above, i_n is just i and all other i's have been condensed into a and b. It would be fantastic if I can work out the inverse function, my thesis is about different data transformations (for which I require the inverse) and it would be extremely useful if I could count a neural network model as a data transformation.

Thanks,

Chewyraver