Ying Bi - Genetic Programming for Image Classification: An Automated Approach to Feature Learning

The role of adaptation, learning and optimization are becoming increasingly essential and intertwined. The capability of a system to adapt either through modification of its physiological structure or via some revalidation process of internal mechanisms that directly dictate the response or behavior is crucial in many real world applications. Optimization lies at the heart of most machine learning approaches while learning and optimization are two primary means to effect adaptation in various forms. They usually involve computational processes incorporated within the system that trigger parametric updating and knowledge or model enhancement, giving rise to progressive improvement. This book series serves as a channel to consolidate work related to topics linked to adaptation, learning and optimization in systems and structures. Topics covered under this series include:

This book series bridges the dichotomy of modern and conventional mathematical and heuristic/meta-heuristics approaches to bring about effective adaptation, learning and optimization. It propels the maxim that the old and the new can come together and be combined synergistically to scale new heights in problem-solving. To reach such a level, numerous research issues will emerge and researchers will find the book series a convenient medium to track the progresses made.

Indexed by SCOPUS, zbMATH, SCImago.

More information about this series at http://www.springer.com/series/8335

This Springer imprint is published by the registered company Springer Nature Switzerland AG

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Commenting on the importance of visual perception, Sternberg and Sternberg write in their textbook on Cognitive Psychology (7th edition, 2017): Vision is the most widely recognized and most widely studied sensory modality. Not only has visual perception received the most interest in the study of sensory functionality in organisms, it turned out to be by far the most important and most complex sense to navigate the world and survive.

As a result of this central importance of vision, the computational and algorithmic aspects of visual perception and vision in computer systems have been at the center of our attempts to create artificially intelligent systems. The consensus is that without a window to the world in the form of sensors, artificial intelligence would be a shadow of its natural counter parts. After the pioneering work of Werner Reichardt on the visual system of flies, David Marrs book Vision: A Computational Investigation into the Human Representation and Processing of Visual Information is the classic to have started the computational era of vision.

Today we live in an age of abundant computational cycles, and the algorithms of the pioneers have been refined and overtaken by more powerful and more versatile algorithms of the GPU era. In fact, computer gaming has a lot to do with this, as it pushed the boundaries of hardware and allowed for massive investment in visual capabilities (in that case for rendering). It turns out, not surprisingly, that the functionalities required for producing images can also be used for processing of images. Thus, GPUs have been recruited for image recognition and image classification and many other image processing tasks. GPUs, it turned out, are even more useful as they provide a computational substrate for neural networks, as they empowered the deep learning revolution in neural network research.

Since the deep learning capabilities have proliferated in computer vision by allowing convolutional neural networks to dominate the scene in Artificial Intelligence (AI), other areas of biologically-inspired computation have also moved forward in their relevance for AI, chief among them Evolutionary Computing. For a long time, already, computational intelligence methods (a subfield of AI concerned with the processing of numerical data) based on evolutionary principles ultimately derived from the natural process of evolution, had been explored. This goes back even to a time before the second waive of neural networks in the 1980s made gradient-based learning algorithms like back-propagation feasible.