Wow! I think this is the first time you've tried it yourself! Thank you so much! How do you think it will converge to a constant? If you send me the file, I'll figure it out quickly. My email address is hjklllllll@gmail.com

If learning is not going well, Please try again by setting the initial weight values. Since we did not set the initial values separately and SGD itself is not an optimizer that learns well, the learning rate can change significantly depending on the initial values. It would be a good idea to try changing the initial values.

Thank you for your immediate reply. I tried setting the initial value several times and it didn't work, so I asked for the last time. I sent it to you by email, so please check it! I followed the implementation of the deep learning model in Excel and it was easy to see and really easy to understand. Even if it's VBA, if you have some programming experience, the syntax is similar, so you can follow it quickly. However, Excel is too slow and freezes often, so that's a bit disappointing. When I go into image CNN, it's very ㅎㅎ; It was a very refreshing lecture. Thank you.

Uh.. I haven't received the email yet.. hjk Next L(l) is 7

There are a few problems 1. The RAND function remains in the X Y data, so the value changes every time 2. The location of the X data in the getdata function is wrong 3. This is not a big deal, but the running rate is too small.. I will correct these points and send it to you again by email.