Model selection: Cross validation

[@Sharon0109]

I think the function with noise should be named as "fn" or any other name, because it's the training data set and the first f is the validation data, we need to make some difference between the two functions.

[@mim8204]

I want the T-shirt.

[@soonmi8278]

Great video

[@amrashraf9516]

I have a question, it's confusing for me why does he compare the coefficients of different parameters as a sign of significance? having the parameters not normalized, so they have significantly different ranges, eg x^1 ranges from [0 - 4], and x^20 ranges from [0 - 1E12]. So I think even if the coefficient of x^20 is small, does not mean that it's ineffective

[@dr.merlot1532]

If after cross validation, you find x^20 coefficient is small, then it is likely not involved in the dynamics of Y=f(X,beta). It will likely not desirable to have an x^20 in your dictionary if you want to do extrapolation of the data. I could be wrong, there are two types of people in this world: Those who is dumb and those who is not. I belong to the former.

[@VIVEKPANDEYIITB]

how do we model a dynamical system with time delay?

[@ErnestoMendoza-oo1fq]

There is some problem with the Python code. The second cell prints a waringin saying: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations, check the scale of the features or consider increasing regularisation. The program still seems to run fine, but still, this is an odd warning. Could you make a video on the Python solution?

[@ErnestoMendoza-oo1fq]

The SVD and the lastq solutions in the Python example do not converge to a parabola as the number of trials increases.

[@ErnestoMendoza-oo1fq]

The two panels showing the Unthresholded and thresholded from Python do not work either.

[@avnerduchovni6675]

Great vid!