Deep learning (Deep Learning) is a machine learning, artificial intelligence and machine learning are essential way.
At present, most outstanding performance of AI applications use a deep learning technology, to lead the third wave of artificial intelligence.
A concept depth study
Depth learning method is characterized learning data based on machine learning.
It belongs to the category of machine learning can be said to be based on traditional neural network upgrade on, approximately equal to the neural network. Its advantage is learning and hierarchical feature extraction algorithm efficient alternative to manually obtain the features of the feature unsupervised or semi-supervised type.
Deep learning is a new field of machine learning 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.
II. Advantages and disadvantages of the depth of learning
Traditional machine learning feature extraction is mainly dependent on manual, time for a specific task simple manual extraction will feature simple and effective, but not universal. Feature extraction depth learning does not rely on manual, but the automatic extraction machines.
☆ The main advantage of the depth of learning are as follows:
Advantage 1: learning ability
From the results, the depth of learning with a strong learning ability, performance is very good.
Advantage 2: wide coverage, adaptability
Neural network learning many layers of depth, width wide, in theory, can be mapped to any function, we are able to solve very complex problems.
Advantage 3: a data driver, a high maximum
Deep learning is highly dependent on the data, the greater the amount of data, the better its performance. In image recognition, face recognition, NLP and other fields in particular.
4 advantages: good portability
Due to the outstanding performance of the depth of learning, a lot of the framework can be used, for example TensorFlow, Pytorch. These frameworks are compatible with many platforms.
☆ depth study is flawed:
One disadvantage: computationally intensive, poor portability
Depth learning requires a lot of data and calculation power, so the cost is very high. And now many applications are not suitable for use on mobile devices. There are already many companies and R & D team in chips for portable devices.
2 shortcomings: high hardware requirements
Depth learning to count forces demanding, ordinary CPU has been unable to meet the requirements of the depth of learning.
3 disadvantages: model design complexity
Model design depth study of very complex and requires a lot of manpower and material resources and time to develop new algorithms and models. Most people only use the ready-made model.
4 drawback: there is no "human nature", easily biased
Since the depth data dependent learning, and interpretability not high. There will be sex discrimination in the training data imbalance circumstances, racial discrimination and so on.
III. 3 typical depth learning algorithm
Convolutional neural network (CNN), recurrent neural network (RNN), generating confrontation Network (GAN) are three typical depth learning algorithm.
Is one of a convolutional neural network-based computing convolutions and having a depth feedforward neural network structure, it is representative of the depth of learning algorithm.
CNN in image processing is very advantageous, currently retrieving the image classification, targeting detection, object segmentation, face recognition, recognition bones and other fields have a wide range of applications.
Neural networks are a class of cyclic recurrent neural network according to the sequence of chained data input, and recursively all of the nodes (circulation means) in the direction of the sequence evolution.
Depth field study, RNN is an effective algorithm for sequence data processing. In text generation, speech recognition, machine translation, generating a field of image is described, the video tag have a wide range of applications.
Generation network against a deep learning model is very popular the last two years of an unsupervised learning algorithm.
GAN can generate a very realistic photos, images and even video, generated image data sets to produce human face photograph, image to image conversion, convert text to images, image editing, picture restoration, and many other fields have a wide range of applications.