what is machine learning

This is really a big question to me, what is exactly machine learning?

Since last year, maybe about Aug 2017, I came to know that there is machine learning, yes, it was that the Computer learns like humans, but at that time I really can’t understand how could it be possible even after watching the course video from Andrew NG. From the video, the basic concept I learn is:

• Machine learning is a proper designed model trained with datasets by gradient decent method to step by step find out the wights and bias.

With this basic concept, I can’t imagine how could it be possible that the objection detection model works so well, why the model looks like that, how the model was designed. I really want to know the reason and why, so for a long time, I am confused, and I don’t know how to go on.

But recent days with some reading papers and knowlege sharing from others on net, it seems that I came to know why, but I don’t know how to say.

For machine learning, it’s really something like our brain, think about that when we are new born, how we learn things step by step, I think this is a combination of supervised learning and unsupervised learning.

But now, most of the models is supervised learning only, such as alextnet, googlenet, MaskRCNN and so on. Then human construct a model, I think it’s something like a new born baby without any knowledges.

So I think if we want to learning machine learning, we should first learn how our branin works, it’s really necessary to know what is a nureon, how it works.

I list below concept as I think I should know if I want to know what is machine learning.

• what is a layer or nureon

• what is activation function

• what is convolution

• what is input

• what is output

• what is neural netwotk