1
$\begingroup$

I am currently using sklearn scale to preprocess my X data before being put into a perceptron - mean/stddev so as to prevent the data converging to infinity or 0. My perceptron returns the weights + bias after the network has been trained:

X = preprocessing.scale(X)

After processing the X and Y data through my perceptron I am returned with weights. From these weights I can calculate the line of best fit:

ls = cp.linspace(cp.min(X), cp.max(X))
best_fit = w[1]+w[0]*ls

where w[1] is a bias. This best_fit line is accurate but it is relative to the preprocessed X rather than the original X which I would like to plot. What is the technique to make these weights relative to the original X values if it is possible?

$\endgroup$

1 Answer 1

1
$\begingroup$

If $$y = w_1 + w_0 \left(\frac{x_{}-\mu_1}{\sigma_1} \right),$$

then we have $$y = \left(w_1-\frac{\mu_1}{\sigma_1}\right)+ \left(\frac{w_0}{\sigma_1}\right)x$$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.