Advantages And Disadvantages Of Deep Learning

Abstract. Deep learning is a new learning algorithm of multi-layer neural network, and it is also a new study field in machine learning. It tries to copy the human brain, which is adept of treating the difficult input data, learning different knowledges intelligently and fast, and solving different kinds of complex problems in a good way. To get these human brain characteristics to a learning model that we want the model with high dimensional data must be compact, support a fast and intelligent learning algorithm and must work with complicated artificial intelligence problems such as computer vision natural language processing and voice recognition. Deep architecture is one of these models, with good learning algorithms for deep learning and

Perceptron is the first machine which can sense and learn but has a primarily partial learning capabilities. The later neural network with multiple hidden layers can learn more complex functions but it has deficiencies of a good learning algorithm. Deep learning is a learning technique with the deep architecture and the good learning algorithms, which can perform the educated learning like learning the features. This ability of deep learning indicate a new direction toward AI. In this survey I will mainly focus on the history, structure, comparisons between different architectures and advantages and disadvantages of deep learning. Section 2 provides a brief history about deep learning. Section 3 introduces the structure of deep learning. Section 4 provides comparisons between different learning architecture. Section 5 provides advantages and disadvantages of deep learning and section 6 will provide the overall conclusion of this survey

Fukushima designed a neural network having multiple of layers and convolutional layers. In 1979, he developed an artificial neural network called neocognitron that used a hierarchical, multilayered design. This project has allowed the computer the "learn" to recognize visual patterns. This Networks was like modern versions, but have designed with a recurrent reinforcement strategy in multiple layers, which has gained strength over time. Furthermore, Fukushima’s design allowed main features to be adjusted manually by increasing the “weight” of certain