What is deep learning?

1、 The difference between machine learning and deep learning
1.1 Machine learning must inform the machine of the characteristics needed to distinguish categories , as follows

Deep learning does not need to inform the machine of the required features , The required features are completely extracted by the neural network , as follows

1.2 Machine learning is absolutely dominant in small data sample analysis ; Deep learning is data driven , Massive data with annotations is the cornerstone of deep learning , Deep learning algorithm is prone to fitting phenomenon in small data sample analysis .

2、 About “ Deep learning ” Various definitions of

2.1 Wu enda ： Deep learning is training neural networks （ Multilayer perceptron (Multi-Layer Perceptron,MLP) Also called Artificial neural network (Artificial Neural Network,ANN), namely In addition to the I / O layer , It can have multiple hidden layers in the middle ）

2.2 Hua Wei Yun ： Deep learning is used to build 、 A neural network that simulates the human brain for analytical learning , A machine learning technique that simulates the mechanism of the human brain to interpret data .
Its basic feature is to try to imitate the transmission between neurons in the brain , Patterns of processing information . The most notable applications are computer vision and natural language processing (NLP) field . obviously ,“ Deep learning ” It is related to machine learning “ neural network ” Is strongly correlated ,“ neural network ” It is also its main algorithm and means ; Or we can put “ Deep learning ” be called “ Improved neural network ” Algorithm . Deep learning and It is divided into convolutional neural network （Convolutional neural networks, abbreviation CNN） And deep confidence network （Deep Belief Nets, abbreviation DBN）. Its main idea is to simulate human neurons , Each neuron receives information , After processing, it can be transmitted to all adjacent neurons . So it looks like the following figure （ Students who want to know more can do it by themselves google）.

3.3 Li Hongyi studies deeply PPT： Deep learning is a kind of neural network ,Deep That means more hidden layers than traditional networks .