Today is a continuation of yesterday, and I tried to make a Neural Network by myself. Click here for yesterday's article
As I wrote in yesterday's article, it's like putting together a number of logistic regressions. We will add more layers to what we have implemented so far.
And the best in Neural Network? The important thing is the activation function.
Since the Neural Network imitates a human nerve cell, it fires when given a value that exceeds a certain standard value. The activation function plays that role.
day4.py The bias is set to be good as a result.
Regarding 1, I think there seems to be a mathematical method. Regarding 2, I think that is what people all over the world are looking for, and I think that there are things that humans can judge and AI can not judge, so if you learn from theory, you will also be interested in such things. I'm glad I had it and started from theory. I will investigate more.
I hope tomorrow I can improve the accuracy a little more or solve either of the questions I wrote earlier. Also, when I look back more, I will write in an easy-to-understand manner how I implemented it.
Recommended Posts