Deep learning is one of the main technologies of machine learning. With deep learning, this chapter is talking about algorithms capable of mimicking the actions of the human brain through artificial neural networks. Compared to other algorithmic structures, neural networks have great advantages: first, their structure based on the stacking of non-linear functions gives them an enormous capacity. In this chapter, the authors will encompass the different models of the convolutional neural network that exist in the literature, their parameters, their architectures, their layers, their advantages, and their disadvantages. The objective is to help researchers to have a summary on deep learning.
TopIntroduction
Deep learning is an ambiguous term, as it has gone through several different meanings throughout the years.
Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator formally to specify all of the knowledge needed by the computer.
Deep Learning is a subset of machine learning, which is in itself a subfield of AI. The figure below is a visual representation of the relationship between AI, ML and DL.
Deep learning, with its remarkable ability to derive meaning from complicated or imprecise data, can be used to extract patterns and detect trends too complex to be noticed by humans or computer techniques. A trained neural network can be considered an “expert” in the category of information it has been given to analyse. This expert can then be used to provide new projections of situations of interest and response. Other benefits include
- 1.
Storage of information on the whole network: information such as in traditional programming is stored on the whole network, not on a database.
- 2.
Ability to work with incomplete knowledge: after training, the data can produce an output even with incomplete information.
- 3.
Fault tolerance: the corruption of one or more cells.
- 4.
Having a distributed memory: for Ann to be able to learn, it is necessary to determine the instances and teach the network according to the desired output by showing these instances to the network. The success of the network is directly proportional to the instances selected and if the event cannot be shown to the network in all its aspects, the network may produce a false output.
- 5.
Ability to make machine learn: Artificial neural networks learn from events and make decisions by commenting on similar events.
- 6.
Parallel processing capability: artificial neural networks have a numerical strength that can perform more than one job at the same time (Jakhar, 2020)
Figure 1.
Relationship between AI, ML and DL (Jakhar, 2020)
The rest of the chapter is structured as following: section two we will present the convolutional neural network family. Section three we will detail the parameters like padding, pooling activation function and edge detection techniques. Section four, illustrates classification models like VGG16, vgg19, AlexNet, GoogleNet ...ect. Section 5 describes the segmentation convolutional model as UNet. Finally, we conclude with a general conclusion and some future works.
TopConvolutional Neural Networks
In recent years, the convolutional neural network (CNN) has emerged as the prevalent model for the machine learning and computer vision. Deep CNNs are widely used in a broad range of real-life applications, such as image classification, object detection and image segmentation. The name “convolutional neural network” indicates that the network employs a mathematical operation called convolution. Convolution is a specialized kind of linear operation. Convolutional networks are simply neural networks that use convolution in place of general matrix multiplication in at least one of their layers. The typical use of convolutional networks is on classification tasks, where the output to an image is a single class label.