Because of the vanishing gradient problem, neural networks lost their popularity in the field of machine learning. We can say that the number of cases used for data mining in the real world by neural networks was remarkably small compared to other typical algorithms such as logistic regression and SVM.
But then deep learning showed up and broke all the existing conventions. As you know, deep learning is the neural network accumulating layers. In other words, it is deep neural networks, and it generates astounding predictability in certain fields. Now, speaking of AI research, it's no exaggeration to say that it's the research into deep neural networks. Surely it's the counterattack by neural networks. If so, why didn't the vanishing gradient problem matter in deep learning? What's the difference between this and the past algorithm?
In this section, we'll look at why deep learning can generate such predictability and its mechanisms.
