Machine intelligence and automation: Deep learning concepts aiding industrial applications

Abstract

Everything begins from a single origin with a unique point in space and time. This is the spark of innovation that fuels the most amazing breakthroughs. It is the passion for discovery that unveils the genesis of all that exists in the universe. Today the power of AI helps computer achieve superhuman capabilities in image recognition and lets scientists save our precious resources by analyzing in 1 month what used to take 10 years. Everyday devices translate even the most complex languages from voice into text and images into words helping the visually impaired recognize an old friend or letting a visually impaired woman read to her child for the first time. Autonomous vehicles give us the freedom to reimagine our city streets where there are no streets, to help the lost to find their way home. We see robots teach themselves to perform simple tasks. We even watch them all as they take their first step. And today a 2500-year-old game meets its match as a computer competes with one of the greatest human champions of all time and wins. This is a collective imagination fuelled by forward-looking technologies and the beginning of the most amazing discoveries yet to come. There are many applications of deep learning. Machine learning provides us an incredible set of tools. If there is a difficult problem in hand, we need not find an algorithm for it, it finds out by itself what is important about the problem and tries to solve it on its own. In many problem domains they perform better than human experts. One of its applications is toxicity detection for different chemical structures by means of deep learning. It is so efficient that it could find the toxic property of the chemicals. Next one is mitosis detection from large images. Mitosis means that certain nuclei are undergoing different transformations that are quite harmful and quite difficult to detect. The technique called convolution neural networks outperforms professional radiologists at their own task. Learning algorithm can read a website and find out whether fake content is discussed there as real content or not. Because of the imperfection of the 3D scanning procedures, 3D scanned furniture is too noisy to use as they are. However, there are techniques to look at these really noisy models and try to figure out how they should look by learning the symmetries and other properties of real furniture. This algorithm can also do an excellent job at predicting how different fluids behave in time and the attire for expected to be super useful in physical simulation in the following years. Another application where a computer algorithm is called deep Q learning plays pong against itself and eventually achieves expertise.