What is the depth of learning
Let's start with the definition of deep learning stuff and the way management at some of the links between deep learning, machine learning, and ai:
Deep learning refers to multi-layered artificial neural network its methods and training. Layer of the neural network will enter a lot number as the matrix, by a non-linear activation Methods weights, and then generate another data set as an output. It's like working mechanism of biological neural brain, like, together, form a neural network "brain" by a suitable matrix quantity, multi-organizational links accurate complex process, just as people recognize objects marked as picture.
Three links as follows:
Machine learning: a path to reach the goal of AI
Deep learning is a machine learning a new field of research, their motivation is to build, simulate the human brain to analyze learning neural network, which mimics the mechanism of the human brain to interpret the data, such as images, sound and text.
Generally speaking, it is to use machine learning algorithms to resolve real data, continuous learning, and then make judgments and predictions of what happened in the world. At this point, researchers do not personally write software to determine the special instruction set, and then let the program perform a specific task, on the contrary, the researchers will use a lot of data and algorithms "training" the machine, the machine learn how to perform tasks.
Machine Learning concept is the early AI researchers have proposed in the past few years, there have been many machine learning algorithm method, including decision tree learning, inductive logic programming, cluster analysis (Clustering), reinforcement learning, Bayesian networks. As you know, nobody really achieve "strong AI" is the ultimate goal, the use of early machine learning methods, our goal even "weak artificial intelligence" is also far from reaching.
In the past many years, the best case application of machine learning is "computer vision", to realize computer vision, researchers still need to manually write a lot of code to complete the task. Manually write the researchers classifier, such as edge detection filter, the only way the program can determine the object where to start, where to end; if the shape of the detected object can be determined there are eight sides; classifier can recognize characters "STOP". Manually written Grouper, researchers can develop a meaningful image recognition algorithms, and then learn to judge, confirms that it is not a stop sign.
Depth learning from machine learning new field developed in artificial neural networks. Early so-called "deep" refers to more than one layer of the neural network. Originally deep learning is not an independent approach to learning, which itself will be used to learn supervised and unsupervised neural network trained depth. However, due to rapid development in recent years in the field, some unique learning tools have been proposed (such as residual network), so more and more people to be treated as a separate method of learning.
The initial depth study is the use of deep neural networks to solve a learning process expression characteristics. Depth is not a new concept of neural network itself, it may be generally understood as a neural network structure comprising a plurality of hidden layers. In order to improve the training effect DNN, people make the appropriate adjustments to aspects such as the connection method and the activation function of neurons.