Hello! I hope you are doing well.
I am practicing fitting the Linear model with random data (other than used in MLS courses). I train two models:
- from scratch, writing code by myself (poor fit)
- copying code from the course (good fit).
The only difference is in a code of gradient descent, attached both. I am wondering why gradient_descent_poor fitted poorly and gradient_descent_good fitted very well. Kindly guide me on this. I will be very thankful to you.
Note: Rest of the code (cost function, gradient, etc.) is the same for both.
Regards,
Saif Ur Rehman.
1 Like
Hi @saifkhanengr ,
You mentioned that all other functions are exactly the same. This, I assume, includes the compute_gradient function.
The first thing I would do is to switch the returning values of the compute_gradient in the _poor version.
Right now you have:
dj_dw, dj_db = compute_gradient…
If you look at the _good version, the returning values are:
dj_db, dj_dw = compute gradient…
Please let me know if this helps.
Thanks,
Juan
IMPORTANT: Please remove code from your question. It is against the honor code of the community to post code.
1 Like
OMG, I made such a silly mistake. Thank you, sir, for highlighting it. Yes, I corrected it and it did work.
The code, in the attached pictures, is from optional labs. If it is still against the honor code, I am extremely sorry for that. I will delete it.
Once again, thank you so much, sir, Juan.
Regards,
Saif Ur Rehman.
1 Like
I am glad to know that this is working now 
Juan
1 Like
